{"draft_name": "draft-niyikiza-oauth-attenuating-agent-tokens", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft defines Attenuating Authorization Tokens (AATs), a JWT-based format that allows cryptographically enforced capability delegation chains for AI agents where each delegation step can only narrow permissions, never expand them. The system extends OAuth Rich Authorization Requests with offline verifiable constraint propagation for tool invocation permissions.\",\n \"n\": 4,\n \"nn\": \"Notable originality in applying cryptographic attenuation principles to AI agent authorization with offline verification of delegation chains\",\n \"m\": 4,\n \"mn\": \"Detailed specification with concrete protocol mechanisms, JWT extensions, and verification algorithms, though implementation examples could be stronger\",\n \"o\": 2,\n \"on\": \"Minor conceptual overlap with general OAuth delegation patterns but unique approach to cryptographic capability attenuation for agents\",\n \"mo\": 2,\n \"mon\": \"Single revision draft with no clear working group adoption signals yet\",\n \"r\": 5,\n \"rn\": \"Directly addresses core AI agent authorization challenges with practical delegation semantics for multi-agent systems\",\n \"c\": [\"Agent identity/auth\", \"A2A protocols\", \"AI safety/alignment\"]\n}", "in_tok": 640, "out_tok": 269} {"draft_name": "draft-araut-oauth-transaction-tokens-for-agents", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft extends OAuth Transaction Tokens to support agent context propagation in multi-agent workflows. It defines mechanisms to identify agents performing actions while preserving principal identity context across agent transitions in complex call chains.\",\n \"n\": 3,\n \"nn\": \"Useful extension of existing OAuth mechanisms for agent-specific use cases, though builds incrementally on established patterns\",\n \"m\": 4,\n \"mn\": \"Well-detailed specification with clear protocol mechanisms, structured claims definitions, and coverage of both single and multi-agent scenarios\",\n \"o\": 3,\n \"on\": \"Shares core concepts with OAuth Transaction Tokens and RFC8693, but applies them specifically to agent contexts\",\n \"mo\": 2,\n \"mon\": \"Single revision with future date suggests early stage, limited evidence of community engagement or working group adoption\",\n \"r\": 5,\n \"rn\": \"Directly addresses core challenges in agent identity and authentication for multi-agent systems and autonomous workflows\",\n \"c\": [\"Agent identity/auth\", \"A2A protocols\"]\n}", "in_tok": 921, "out_tok": 247} {"draft_name": "draft-xsaopig-nmop-service-flow-modal-mapping", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft proposes a framework for mapping service flow characteristics to network resources in multi-modal intelligent computing networks using ALTO protocol and SDN architecture. It focuses on dynamic resource allocation and network optimization through intelligent mapping techniques.\",\n \"n\": 2,\n \"nn\": \"Incremental work combining existing ALTO protocol with SDN for service flow mapping - straightforward application of known technologies\",\n \"m\": 2,\n \"mn\": \"Early architectural sketch without detailed protocol specifications, implementation details, or concrete examples\",\n \"o\": 4,\n \"on\": \"Significant overlap with existing SDN traffic engineering and ALTO-based network optimization work\",\n \"mo\": 1,\n \"mon\": \"No evidence of working group adoption or community engagement, appears to be individual submission\",\n \"r\": 3,\n \"rn\": \"Partially relevant - addresses intelligent network management but focuses more on traditional network optimization than AI agents\",\n \"c\": [\"ML traffic mgmt\", \"Autonomous netops\"]\n}", "in_tok": 650, "out_tok": 236} {"draft_name": "draft-rosenberg-aiproto-a2t", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft defines the Agent-to-Tool (A2T) Protocol that standardizes how AI agents discover and invoke third-party tools/APIs. It provides enumeration capabilities for tool discovery and invocation mechanisms for runtime execution.\",\n \"n\": 3,\n \"nn\": \"Useful contribution addressing practical AI agent integration needs, though tool invocation patterns are becoming established\",\n \"m\": 3,\n \"mn\": \"Defined protocol with clear enumeration and invocation APIs, but lacks detailed implementation examples and edge case handling\",\n \"o\": 3,\n \"on\": \"Shares conceptual space with OpenAI function calling specs and general API integration patterns, but focuses on standardized discovery\",\n \"mo\": 2,\n \"mon\": \"Single revision from 2026 with no clear working group adoption or community feedback visible\",\n \"r\": 5,\n \"rn\": \"Core AI agent functionality - tool integration is fundamental to practical agent deployment in enterprise environments\",\n \"c\": [\"A2A protocols\", \"Data formats/interop\", \"Agent discovery/reg\"]\n}", "in_tok": 702, "out_tok": 255} {"draft_name": "draft-zm-rtgwg-mcp-troubleshooting", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft proposes using the Model Context Protocol (MCP) to enable network devices to act as MCP servers and controllers as clients for LLM-based network troubleshooting. It extends the model to support device-to-device collaboration for distributed fault correlation when controllers are unreachable.\",\"n\":3,\"nn\":\"Useful application of existing MCP standard to network domain with reasonable D2D extension\",\"m\":3,\"mn\":\"Defines protocol mapping and architecture but lacks detailed implementation specifics\",\"o\":2,\"on\":\"Some conceptual overlap with intent-based networking but novel MCP application\",\"mo\":2,\"mon\":\"Single revision with future date suggests early stage development\",\"r\":4,\"rn\":\"Directly addresses AI-driven network automation with LLM integration\",\"c\":[\"Autonomous netops\",\"A2A protocols\",\"Data formats/interop\"]}", "in_tok": 673, "out_tok": 190} {"draft_name": "draft-ietf-lisp-nexagon", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"Lisp-Nexagon combines LISP with H3 spatial indexing to create a geospatial intelligence network for distributed agents in transportation, public safety, and logistics applications. The protocol enables agents to report, aggregate, and disseminate location-based state information across physical-world systems.\",\"n\":4,\"nn\":\"Novel combination of LISP networking with H3 geospatial indexing for agent-based geo-intelligence - creative architectural fusion\",\"m\":3,\"mn\":\"Defined protocol concept with clear architecture but likely needs more implementation details and examples\",\"o\":2,\"on\":\"Some conceptual overlap with location-based networking but the H3+LISP combination appears relatively unique\",\"mo\":3,\"mon\":\"Recent draft suggesting active development, though unclear about broader community engagement at this stage\",\"r\":4,\"rn\":\"Directly relevant to AI agents operating in physical world scenarios with geospatial coordination needs\",\"c\":[\"A2A protocols\",\"Data formats/interop\",\"Agent discovery/reg\"]}", "in_tok": 534, "out_tok": 224} {"draft_name": "draft-teodor-pilot-problem-statement", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft proposes network-layer infrastructure to give AI agents first-class network citizenship, moving beyond current application-layer HTTP-based communication. It identifies the need for agent-specific addressing, identity, and transport mechanisms independent of web infrastructure designed for humans.\",\n \"n\": 4,\n \"nn\": \"Notable originality in recognizing agents as a distinct network participant class requiring dedicated network-layer support rather than just application protocols\",\n \"m\": 1,\n \"mn\": \"Problem statement only with requirements identification but no technical solutions or protocol mechanisms defined\",\n \"o\": 2,\n \"on\": \"Minor conceptual similarities to general identity/addressing work but unique focus on agent-specific network layer needs\",\n \"mo\": 2,\n \"mon\": \"Single revision with future date suggests early stage without demonstrated community engagement or WG activity\",\n \"r\": 5,\n \"rn\": \"Core AI agent topic addressing fundamental networking infrastructure for autonomous agent communication\",\n \"c\": [\"A2A protocols\", \"Agent identity/auth\", \"Agent discovery/reg\"]\n}", "in_tok": 575, "out_tok": 249} {"draft_name": "draft-wang-nmrg-magent-im", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft proposes a protocol-agnostic data model for multi-agent communication in Level 4 autonomous networks, enabling intelligent agents to interact securely and efficiently for self-management and healing. The model aims to support end-to-end task execution and closed-loop operations across network services and resource layers.\",\n \"n\": 3,\n \"nn\": \"Useful contribution addressing multi-agent coordination in autonomous networks, though builds on established autonomous networking concepts\",\n \"m\": 2,\n \"mn\": \"Early sketch stage - presents the problem and high-level model concept but lacks detailed protocol specifications or implementation details\",\n \"o\": 3,\n \"on\": \"Shares conceptual overlap with existing autonomous network management and multi-agent system drafts, but focuses specifically on protocol-agnostic interaction models\",\n \"mo\": 2,\n \"mon\": \"Single revision from January 2026 with no clear working group adoption or community engagement visible\",\n \"r\": 5,\n \"rn\": \"Core AI agent topic directly addressing multi-agent interaction models for autonomous network operations\",\n \"c\": [\"A2A protocols\", \"Autonomous netops\", \"Data formats/interop\"]\n}", "in_tok": 647, "out_tok": 279} {"draft_name": "draft-cui-nmrg-llm-benchmark", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"Proposes a comprehensive evaluation framework for LLM-based network configuration agents, combining emulator environments, representative tasks, and multi-dimensional metrics. The framework aims to enable reproducible comparisons of intent-driven network configuration approaches using standardized assessment criteria.\",\n \"n\": 3,\n \"nn\": \"Useful contribution addressing the gap in standardized evaluation methods for LLM network agents, though evaluation frameworks are a natural extension of existing ML benchmarking practices\",\n \"m\": 4,\n \"mn\": \"Detailed specification with concrete framework components, task definitions, and evaluation metrics, appears implementation-ready with clear technical structure\",\n \"o\": 2,\n \"on\": \"Minor similarities to general ML evaluation practices but unique focus on network configuration domain with specialized metrics and emulator integration\",\n \"mo\": 2,\n \"mon\": \"Single revision draft with recent publication date, no clear indication of working group adoption or broader community engagement yet\",\n \"r\": 5,\n \"rn\": \"Core AI agent topic directly addressing evaluation and benchmarking of LLM agents in network operations context\",\n \"c\": [\"Autonomous netops\", \"Data formats/interop\"]\n}", "in_tok": 554, "out_tok": 275} {"draft_name": "draft-westerbeck-reason-protocol", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"Defines a reason:// URI scheme and registry protocol for storing and retrieving compressed reasoning artifacts from autonomous agents. Artifacts enter the registry through competitive arbitration scored by the Web Agent Reasoning Federation (WARF) protocol, creating DNS-like infrastructure for the agentic internet.\",\n \"n\": 5,\n \"nn\": \"Genuinely novel approach combining URI schemes, competitive arbitration, and structural reasoning artifacts with mathematical non-invertibility - no existing work addresses this specific combination\",\n \"m\": 2,\n \"mn\": \"Early sketch stage - describes high-level concept and competitive mechanism but lacks detailed protocol specification, message formats, or implementation guidance\",\n \"o\": 1,\n \"on\": \"Unique approach - while URI schemes and registries exist, the combination with competitive reasoning artifact arbitration and WARF scoring is distinctive\",\n \"mo\": 2,\n \"mon\": \"Single revision with future date (2026-03-30) suggests early conceptual work without demonstrated community engagement or working group adoption\",\n \"r\": 5,\n \"rn\": \"Core AI agent topic directly addressing infrastructure needs for distributed autonomous agent networks and reasoning artifact management\",\n \"c\": [\"A2A protocols\", \"Agent discovery/reg\", \"Data formats/interop\"]\n}", "in_tok": 601, "out_tok": 296} {"draft_name": "draft-altanai-aipref-realtime-protocol-bindings", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft defines protocol bindings for AI preferences in real-time communications, specifying how SIP and SDP can carry AI processing requirements and constraints. It provides concrete header fields and attributes for endpoints and AI assistants to negotiate AI-driven processing of session data.\",\"n\":4,\"nn\":\"Notable approach to integrate AI preferences directly into established real-time protocols like SIP/SDP, addressing a gap in how AI capabilities are signaled in communications\",\"m\":3,\"mn\":\"Defines specific protocol mechanisms and binding models with concrete SIP headers and SDP attributes, though implementation details and examples could be more comprehensive\",\"o\":2,\"on\":\"While builds on existing SIP/SDP foundations, the AI preference binding approach appears relatively unique with minor conceptual overlap to other AI signaling work\",\"mo\":2,\"mon\":\"Single revision from February 2026 with no clear indication of working group adoption or broader community engagement yet\",\"r\":5,\"rn\":\"Core AI agent topic directly addressing how AI capabilities and preferences are signaled and negotiated in real-time communication protocols\",\"c\":[\"A2A protocols\",\"Data formats/interop\",\"Agent discovery/reg\"]}", "in_tok": 620, "out_tok": 250} {"draft_name": "draft-intellinode-cats-in-network-scheduling", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"IntelliNode proposes an in-network intelligent scheduling mechanism for CATS that uses FPGAs and programmable switches to perform real-time scheduling decisions for AI workloads. It creates a closed-loop system that extracts features at line rate and makes heuristic decisions for path selection, tensor slicing, and compute matching.\",\n \"n\": 4,\n \"nn\": \"Notable originality in combining FPGA-based in-network processing with AI workload-aware scheduling, moving beyond traditional probe-based approaches\",\n \"m\": 2,\n \"mn\": \"Early sketch with high-level architecture description but lacks detailed protocol specifications, message formats, or concrete implementation details\",\n \"o\": 2,\n \"on\": \"Minor similarities to CATS framework extensions but introduces unique FPGA-based in-network intelligence approach not seen in other drafts\",\n \"mo\": 2,\n \"mn\": \"Single revision from 2026 with no indication of working group adoption or community feedback yet\",\n \"r\": 4,\n \"rn\": \"Directly relevant to AI agent networking as it addresses intelligent traffic steering and scheduling for distributed AI training/inference workloads\",\n \"c\": [\"ML traffic mgmt\", \"Autonomous netops\"]\n}", "in_tok": 736, "out_tok": 294} {"draft_name": "draft-song-rtgwg-din-usecases-requirements", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft presents problem statement, use cases, and requirements for Distributed Inference Network (DIN) architecture to address scalability, latency, and security challenges of centralized AI inference through edge-cloud collaboration. It focuses on enabling low-latency, high-concurrency AI inference services across distributed network infrastructure.\",\n \"n\": 2,\n \"nn\": \"Incremental contribution applying well-known edge computing and CDN distribution concepts to AI inference workloads, without novel architectural insights\",\n \"m\": 1,\n \"mn\": \"Problem statement phase only, lacking concrete technical mechanisms, protocols, or implementation details for the proposed DIN architecture\",\n \"o\": 4,\n \"on\": \"Significant overlap with existing edge computing, mobile edge computing (MEC), and distributed AI serving literature and standards work\",\n \"mo\": 2,\n \"mon\": \"Single revision draft with no clear working group adoption or broader community engagement visible\",\n \"r\": 4,\n \"rn\": \"Directly addresses AI inference infrastructure challenges, which is highly relevant to AI agent deployment and scalability requirements\",\n \"c\": [\"Model serving/inference\", \"ML traffic mgmt\", \"Autonomous netops\"]\n}", "in_tok": 593, "out_tok": 280} {"draft_name": "draft-liu-rtgwg-mdt-in-high-bdp", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft proposes a Massive Data Transmission (MDT) framework for efficiently utilizing nighttime idle bandwidth in high BDP networks through terminal-driven intelligent parameter optimization. The system adjusts local configurations like disk I/O, NIC bandwidth, and TCP buffers to match transmission goals for services like same-day delivery.\",\"n\":2,\"nn\":\"Incremental improvement to existing bandwidth optimization and traffic management techniques, applying known concepts to nighttime idle periods\",\"m\":2,\"mn\":\"Early sketch with high-level architecture described but lacking detailed protocol specifications, implementation details, or concrete mechanisms\",\"o\":3,\"on\":\"Shares traffic management and QoS optimization concepts with existing network management drafts, though specific focus on nighttime bandwidth utilization is somewhat distinct\",\"mo\":2,\"mon\":\"Single revision draft with no evidence of working group adoption or broader community engagement\",\"r\":2,\"rn\":\"Tangentially related through mentions of 'intelligent parameter optimization' but primarily focuses on traditional network traffic management rather than AI agents\",\"c\":[\"ML traffic mgmt\"]}", "in_tok": 636, "out_tok": 228} {"draft_name": "draft-fluid-fadp", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"FADP defines an HTTP extension that allows AI agents to pay for web resources using cryptocurrency micropayments, embedding cryptographic payment proofs in HTTP headers. The protocol extends HTTP 402 with new headers for payment requirements and verification, targeting sub-dollar transactions where traditional billing is impractical.\",\n \"n\": 4,\n \"nn\": \"Notable integration of blockchain micropayments with HTTP for autonomous agents - novel approach to solving agent payment authentication\",\n \"m\": 4,\n \"mn\": \"Detailed specification with concrete HTTP header definitions, nonce mechanisms, and verification flows - appears implementation-ready\",\n \"o\": 2,\n \"on\": \"Some conceptual similarity to HTTP payment extensions but unique in blockchain integration and agent-specific design\",\n \"mo\": 2,\n \"mon\": \"Single draft submission with no clear working group adoption or community engagement indicators\",\n \"r\": 5,\n \"rn\": \"Directly addresses core autonomous agent needs for resource access and payment authentication in decentralized scenarios\",\n \"c\": [\"Agent identity/auth\", \"A2A protocols\"]\n}", "in_tok": 742, "out_tok": 256} {"draft_name": "draft-crovia-seal", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"The Crovia Seal defines a cryptographic receipt format using Ed25519 signatures and hash chains to provide tamper-evident provenance records for AI-generated content. The specification includes optional transparency log integration and witness co-signatures for enhanced verification.\",\n \"n\": 3,\n \"nn\": \"Useful contribution applying established cryptographic primitives (Ed25519, hash chains, transparency logs) to the emerging need for AI output provenance verification\",\n \"m\": 4,\n \"mn\": \"Detailed specification with specific cryptographic choices (Ed25519, SHA-256, CSC-1 canonicalization), clear field definitions, and considerations for offline verification and third-party integration\",\n \"o\": 2,\n \"on\": \"Minor similarities to general digital signature and transparency log approaches, but specific application to AI provenance appears relatively unique\",\n \"mo\": 2,\n \"mon\": \"Single revision with no clear indication of working group adoption or broader community engagement yet\",\n \"r\": 4,\n \"rn\": \"Directly addresses AI output authenticity and provenance, which are increasingly critical concerns for AI systems deployment and trustworthiness\",\n \"c\": [\"Data formats/interop\", \"AI safety/alignment\", \"Agent identity/auth\"]\n}", "in_tok": 660, "out_tok": 295} {"draft_name": "draft-song-oauth-ai-agent-authorization", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft proposes adding a 'target_id' field to OAuth 2.0 to enable granular authorization for AI agents. The extension aims to provide precise permission management and enhance traceability while maintaining compatibility with existing OAuth workflows.\",\n \"n\": 2,\n \"nn\": \"Adding a target identifier to OAuth is an incremental extension of existing authorization patterns, not a fundamentally new approach\",\n \"m\": 2,\n \"mn\": \"Abstract suggests early-stage work with basic concept outlined but lacks detailed protocol specification or implementation guidance\",\n \"o\": 3,\n \"on\": \"Shares authorization concepts with existing OAuth extensions and fine-grained access control mechanisms, though AI agent focus is somewhat distinct\",\n \"mo\": 2,\n \"mon\": \"Appears to be early draft from 2026 with limited indication of broader community engagement or working group adoption\",\n \"r\": 4,\n \"rn\": \"Directly addresses AI agent authorization challenges, which is highly relevant to autonomous systems requiring controlled access to resources\",\n \"c\": [\"Agent identity/auth\", \"Policy/governance\"]\n}", "in_tok": 601, "out_tok": 260} {"draft_name": "draft-gupta-httpapi-events-query", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"HTTP Events Query Protocol (EQP) enables user agents to receive event notifications directly from HTTP resources using a minimal protocol extension. The draft proposes that resources themselves should be the intuitive source for event notifications rather than separate event systems.\",\"n\":2,\"nn\":\"Incremental approach to HTTP-based event delivery, building on existing patterns but not fundamentally novel\",\"m\":4,\"mn\":\"Detailed 29-page specification with clear protocol definitions and examples, appears implementation-ready\",\"o\":4,\"on\":\"Significant overlap with Server-Sent Events, WebSockets, and other HTTP event streaming mechanisms\",\"mo\":2,\"mon\":\"Single revision draft with October 2025 date, no evidence of working group adoption or community engagement\",\"r\":3,\"rn\":\"Partially relevant as it addresses agent-to-resource communication patterns, though not specifically AI-focused\",\"c\":[\"A2A protocols\",\"Data formats/interop\"]}", "in_tok": 504, "out_tok": 204} {"draft_name": "draft-morrison-mcp-tool-surface-names-registry", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft establishes an IANA registry for Model Context Protocol (MCP) tool surface names, which are wire-level identifiers used by clients to invoke typed capabilities on MCP servers. It defines a Specification Required registration process with Designated Expert review and includes initial entries from related Morrison-family drafts.\",\"n\":2,\"nn\":\"Registry establishment is incremental work building on existing MCP specifications, following standard IANA patterns\",\"m\":3,\"mn\":\"Defines clear registry structure, registration procedures, and expert review process but lacks implementation examples\",\"o\":2,\"on\":\"Builds on Morrison-family MCP drafts but represents distinct registry infrastructure component\",\"mo\":3,\"mon\":\"Active development as part of broader MCP specification suite, though limited to registry mechanics\",\"r\":4,\"rn\":\"Directly supports AI agent infrastructure by standardizing tool capability identifiers in MCP ecosystem\",\"c\":[\"Agent discovery/reg\",\"Data formats/interop\",\"A2A protocols\"]}", "in_tok": 636, "out_tok": 213} {"draft_name": "draft-an-nmrg-i2icf-cits", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft specifies a framework for orchestrating In-Network Computing Functions (ICFs) in Cooperative Intelligent Transportation Systems, enabling vehicles, roadside units, and network domains to collaborate through Agent-to-Agent communication. The approach aims to optimize real-time V2X communications, traffic management, and edge computing services through intelligent agent coordination.\",\n \"n\": 3,\n \"nn\": \"Useful contribution combining in-network computing with C-ITS through A2A paradigms, though individual concepts exist separately\",\n \"m\": 2,\n \"mn\": \"Early sketch level - abstract describes framework goals but lacks concrete protocol details or implementation specifics\",\n \"o\": 3,\n \"on\": \"Shares concepts with edge computing drafts and vehicular networking work, but specific C-ITS+ICF+A2A combination has limited overlap\",\n \"mo\": 2,\n \"mon\": \"Single revision in NMRG, limited visibility of broader community engagement or working group momentum\",\n \"r\": 4,\n \"rn\": \"Directly relevant with explicit Agent-to-Agent communication paradigm and intelligent agents coordinating computing tasks\",\n \"c\": [\"A2A protocols\", \"ML traffic mgmt\", \"Autonomous netops\", \"Agent discovery/reg\"]\n}", "in_tok": 675, "out_tok": 299} {"draft_name": "draft-li-cats-idn", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft proposes an Intelligence Delivery Network (IDN) framework for distributing deep learning inference models across geographically distributed servers to optimize latency, privacy, and scalability. The document presents architectural concepts and vocabulary but does not specify actual protocols or implementation details.\",\"n\":2,\"nn\":\"Incremental extension of CDN concepts to AI inference workloads, builds on established distributed computing patterns\",\"m\":2,\"mn\":\"Early architectural sketch with framework concepts but lacks concrete protocols, mechanisms, or detailed specifications\",\"o\":3,\"on\":\"Shares distributed inference concepts with some existing work in edge AI and model serving architectures\",\"mo\":2,\"mon\":\"Single draft submission with no indication of working group adoption or community development activity\",\"r\":5,\"rn\":\"Directly addresses core AI infrastructure challenges around distributed model serving and inference optimization\",\"c\":[\"Model serving/inference\",\"ML traffic mgmt\"]}", "in_tok": 588, "out_tok": 194} {"draft_name": "draft-aevum-wlp", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"WLP v0 defines a protocol for translating AI-generated scientific hypotheses into machine-executable wet-lab automation instructions, specifically targeting longevity research applications. The draft provides a royalty-free schema with reference implementations in Rust and Python for AI-to-automation instruction translation.\",\n \"n\": 4,\n \"nn\": \"Notable originality in bridging AI hypothesis generation with physical lab automation through structured protocols - addresses a genuinely important gap in scientific automation workflows\",\n \"m\": 3,\n \"mn\": \"Defined protocol with schema specification and reference implementations mentioned, though lacks detailed examples or test vectors in the abstract\",\n \"o\": 2,\n \"on\": \"References SCP from photoresist domain but appears to be novel application to wet-lab/longevity research with minimal overlap to existing IETF work\",\n \"mo\": 2,\n \"mon\": \"Single revision draft with no clear indication of working group adoption or broader community engagement yet\",\n \"r\": 4,\n \"rn\": \"Directly relevant to AI agent systems - specifically addresses agent-to-automation interfaces and structured instruction protocols for physical world interaction\",\n \"c\": [\"A2A protocols\", \"Data formats/interop\", \"Human-agent interaction\"]\n}", "in_tok": 584, "out_tok": 294} {"draft_name": "draft-zhang-rvp-problem-statement", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"RVP proposes a protocol for coordinating physical entities (robots, IoT devices) with digital agents (AI systems, virtual twins) through unified identity management and graph-based coordination. The draft addresses the gap between existing peer-to-peer agent protocols and the hierarchical, constraint-aware communication needed for embodied intelligence in manufacturing, smart cities, and autonomous systems.\",\n \"n\": 4,\n \"nn\": \"Notable originality in bridging physical-digital agent coordination with explicit physical constraint integration and non-peer hierarchical relationships\",\n \"m\": 1,\n \"mn\": \"Problem statement only - no protocol specification, message formats, or technical mechanisms defined\",\n \"o\": 2,\n \"on\": \"Minor similarities to existing agent protocols (A2A, MCP, ANP) but addresses a distinct physical-digital integration gap\",\n \"mo\": 2,\n \"mon\": \"Single draft submission with no evidence of working group adoption or community engagement\",\n \"r\": 5,\n \"rn\": \"Core AI agent topic addressing critical embodied intelligence coordination challenges with direct relevance to physical-world AI deployment\",\n \"c\": [\"A2A protocols\", \"Agent identity/auth\", \"Data formats/interop\", \"Human-agent interaction\"]\n}", "in_tok": 678, "out_tok": 293} {"draft_name": "draft-wang-hjs-accountability", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"HJS defines an accountability layer for AI agents using four primitives (Judgment/Delegation/Termination/Verification) to record responsibility chains. The system leverages Bitcoin blockchain anchoring via Open Timestamp proofs to create immutable, court-admissible evidence of AI agent actions.\",\n \"n\": 4,\n \"nn\": \"Notable originality in combining blockchain-anchored timestamping with AI agent accountability primitives, creating a novel approach to legal evidence generation for AI systems\",\n \"m\": 2,\n \"mn\": \"Early sketch stage - defines core primitives and blockchain integration concept but lacks detailed protocol specification, message formats, and implementation guidance\",\n \"o\": 2,\n \"on\": \"Minor similarities to other accountability frameworks but unique in its specific focus on blockchain-anchored legal evidence and the four-primitive accountability model\",\n \"mo\": 2,\n \"mon\": \"Single revision with no indication of working group adoption or broader community engagement in AI accountability standards\",\n \"r\": 5,\n \"rn\": \"Core AI agent topic addressing critical accountability and legal responsibility challenges that are fundamental to AI agent deployment\",\n \"c\": [\"AI safety/alignment\", \"Policy/governance\", \"Human-agent interaction\"]\n}", "in_tok": 565, "out_tok": 287} {"draft_name": "draft-tonyai-a2a-trust", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft defines a cryptographic trust model for agent-to-agent interactions using PKI and OAuth patterns to establish verifiable identities, spawn chains, and cross-organizational authorization. It applies existing security primitives to the specific problem of agent provenance in multi-agent AI systems.\",\"n\":3,\"nn\":\"Useful contribution applying established PKI/OAuth patterns to agent trust - not novel cryptographically but addresses a real emerging need\",\"m\":3,\"mn\":\"Defines clear protocol mechanisms with CA templates and dual-signature policies, but lacks detailed examples and implementation specifics\",\"o\":2,\"on\":\"While identity and trust are common themes, the specific A2A spawn chain and template approach has minor similarities to existing work\",\"mo\":2,\"mon\":\"Single revision from 2026 with no evidence of working group activity or community engagement yet\",\"r\":5,\"rn\":\"Directly addresses core AI agent infrastructure - identity, trust, and provenance are fundamental for multi-agent systems\",\"c\":[\"Agent identity/auth\",\"A2A protocols\",\"Policy/governance\"]}", "in_tok": 641, "out_tok": 230} {"draft_name": "draft-jeong-opsawg-intent-based-sdv-framework", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft proposes an intent-based management framework for Software-Defined Vehicles (SDVs) to enable autonomous vehicle communication and coordination in intelligent transportation systems. The framework aims to simplify network, security, and application configuration in SDVs through high-level intent specifications rather than low-level device configuration.\",\n \"n\": 2,\n \"nn\": \"Incremental application of well-established intent-based networking concepts to the automotive domain, without significant technical innovation beyond domain adaptation\",\n \"m\": 2,\n \"mn\": \"Early architectural sketch with high-level framework description but lacks detailed protocol specifications, implementation details, or concrete examples\",\n \"o\": 4,\n \"on\": \"Significant conceptual overlap with existing intent-based networking frameworks and SDN management approaches, primarily distinguished by automotive use case focus\",\n \"mo\": 2,\n \"mon\": \"Appears to be single revision with no indication of working group adoption or broader community engagement in IETF context\",\n \"r\": 3,\n \"rn\": \"Partially relevant as SDVs involve autonomous agents requiring coordination, but focuses more on network management infrastructure than core AI agent behaviors\",\n \"c\": [\"Autonomous netops\", \"A2A protocols\"]\n}", "in_tok": 610, "out_tok": 289} {"draft_name": "draft-aap-oauth-profile", "model": "claude-sonnet-4-20250514", "raw": "```json\n{\n \"s\": \"This draft defines an authorization profile for OAuth 2.0 specifically designed for autonomous AI agents, extending existing standards with structured claims for agent identity, task context, and oversight requirements. It focuses on enabling context-aware, auditable authorization in agent-to-API scenarios without introducing new protocols.\",\n \"n\": 4,\n \"nn\": \"Notable originality in adapting OAuth 2.0 for autonomous agents with structured claims for delegation chains and human oversight - addresses genuinely new challenges\",\n \"m\": 4,\n \"mn\": \"At 83 pages, appears to be a detailed specification with comprehensive coverage of agent authorization scenarios and structured claims validation\",\n \"o\": 2,\n \"on\": \"While building on OAuth 2.0 and JWT, the agent-specific extensions and context-aware authorization mechanisms represent a fairly unique approach\",\n \"mo\": 3,\n \"mon\": \"Recent draft (2026-02-07) suggesting active development, though momentum level unclear without revision history\",\n \"r\": 5,\n \"rn\": \"Core AI agent topic - directly addresses critical need for agent identity, authorization, and oversight in autonomous systems\",\n \"c\": [\"Agent identity/auth\", \"AI safety/alignment\", \"Policy/governance\"]\n}\n```", "in_tok": 582, "out_tok": 297} {"draft_name": "draft-hori-agent-quality-graph", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"Proposes Agent Quality Graph (AQG) protocol for evaluating AI agent trustworthiness through delegation transaction graphs, similar to PageRank for web pages. Defines delegation record formats, graph construction, scoring algorithms, and query APIs for trust scores.\",\"n\":4,\"nn\":\"Notable adaptation of PageRank-style algorithms to agent trustworthiness - creative cross-domain application\",\"m\":3,\"mn\":\"Defines core protocol components and algorithms but lacks detailed implementation specifics\",\"o\":2,\"on\":\"Some conceptual overlap with reputation systems but unique application to agent delegation graphs\",\"mo\":2,\"mon\":\"Single revision with future date suggests early stage work\",\"r\":5,\"rn\":\"Directly addresses core challenge of agent trustworthiness evaluation in multi-agent systems\",\"c\":[\"A2A protocols\",\"AI safety/alignment\",\"Agent identity/auth\"]}", "in_tok": 623, "out_tok": 190} {"draft_name": "draft-xie-ai-agent-multimodal", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft specifies multimodal requirements for agent-to-agent protocols, enabling autonomous agents to negotiate heterogeneous data capabilities and exchange synchronized content across text, file, audio/video, and sensor streams. It focuses on multi-channel communication sessions with adaptive QoS policies for diverse data types.\",\"n\":3,\"nn\":\"Useful contribution addressing multimodal agent communication, though builds on established multimedia protocol concepts\",\"m\":2,\"mn\":\"Early sketch stage - appears to be requirements document without detailed protocol mechanisms or implementation details\",\"o\":3,\"on\":\"Shares concepts with existing multimedia protocols like WebRTC and SIP, but applies them specifically to autonomous agent communication\",\"mo\":2,\"mon\":\"Single revision with recent publication date, no evidence of working group adoption or community engagement yet\",\"r\":5,\"rn\":\"Core AI agent topic directly addressing autonomous agent communication protocols and multimodal data exchange\",\"c\":[\"A2A protocols\",\"Data formats/interop\"]}", "in_tok": 529, "out_tok": 209} {"draft_name": "draft-cui-dmsc-agent-cdi", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"Defines a comprehensive framework for AI agents to collaborate across different administrative domains and networks. Addresses identity federation, trust establishment, policy harmonization, and secure communication for cross-domain AI agent interoperability.\",\"n\":4,\"nn\":\"Notable originality in addressing cross-domain AI agent collaboration challenges with federated approach\",\"m\":3,\"mn\":\"Defines clear protocol mechanisms and framework components but lacks detailed implementation specifics\",\"o\":2,\"on\":\"Some conceptual overlap with general federation protocols but unique focus on AI agent collaboration\",\"mo\":2,\"mon\":\"Single revision draft with no clear working group adoption yet\",\"r\":5,\"rn\":\"Core AI agent topic addressing fundamental interoperability challenges for multi-domain collaboration\",\"c\":[\"A2A protocols\",\"Agent identity/auth\",\"Policy/governance\",\"Agent discovery/reg\"]}", "in_tok": 575, "out_tok": 178} {"draft_name": "draft-vinaysingh-awp-wellknown", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft proposes a well-known URI (/.well-known/awp.json) and link relation to expose machine-readable descriptions of website workflows, enabling automated agents to interact predictably without web scraping. The format uses minimal JSON to describe states, actions, and events for common website operations.\",\n \"n\": 4,\n \"nn\": \"Notable originality in creating a standardized way for websites to expose workflow metadata specifically for agent consumption, moving beyond traditional API approaches\",\n \"m\": 3,\n \"mn\": \"Defines the protocol mechanism and JSON format but lacks detailed implementation examples and comprehensive specification of workflow description schemas\",\n \"o\": 2,\n \"on\": \"Some conceptual similarity to web discovery mechanisms like robots.txt or OpenAPI, but unique focus on workflow state machines for agents\",\n \"mo\": 2,\n \"mon\": \"Appears to be early-stage individual submission without clear working group adoption or broader community engagement\",\n \"r\": 5,\n \"rn\": \"Directly addresses core challenge of agent-web interaction by providing structured alternative to web scraping, highly relevant to autonomous agent operations\",\n \"c\": [\"Agent discovery/reg\", \"Data formats/interop\", \"A2A protocols\"]\n}", "in_tok": 543, "out_tok": 288} {"draft_name": "draft-ietf-dnsop-dnssec-bootstrapping", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This IETF draft proposes an in-band method for DNS operators to publish authenticated DNSSEC key parameters, enabling automatic bootstrapping of DNSSEC without out-of-band validation. The mechanism allows parent registries to cryptographically validate CDS/CDNSKEY records even when DS records are absent.\",\n \"n\": 3,\n \"nn\": \"Addresses real DNSSEC deployment friction with a useful cryptographic validation approach, though builds incrementally on existing CDS/CDNSKEY mechanisms\",\n \"m\": 4,\n \"mn\": \"Well-developed specification with clear protocol mechanics, validation procedures, and security considerations, approaching implementation-ready status\",\n \"o\": 3,\n \"on\": \"Shares core concepts with RFC 8078 and RFC 7344 on DNSSEC automation, but proposes a distinct in-band validation approach\",\n \"mo\": 4,\n \"mon\": \"IETF working group draft with active GitHub collaboration and clear preference establishment over existing methods indicates strong momentum\",\n \"r\": 1,\n \"rn\": \"Pure DNS infrastructure protocol focused on DNSSEC bootstrapping with no AI or agent-related functionality\",\n \"c\": [\"Other AI/agent\"]\n}", "in_tok": 770, "out_tok": 294} {"draft_name": "draft-rosenberg-oauth-aauth", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft defines an OAuth 2.1 extension for AI agents to obtain access tokens using PII collected through natural language conversations over channels like phone/text that don't support traditional OAuth flows. The protocol aims to prevent LLM hallucinations from causing impersonation attacks while enabling agent-based API access.\",\n \"n\": 4,\n \"nn\": \"Novel approach to OAuth for conversational AI agents using PII-based authentication over non-web channels, addressing unique security challenges of LLM-powered systems\",\n \"m\": 2,\n \"mn\": \"Early sketch stage - defines concept and use case but lacks detailed protocol mechanisms, security analysis, and implementation specifics\",\n \"o\": 2,\n \"on\": \"Minor similarities to existing OAuth extensions but addresses unique problem space of conversational AI agents with distinct security considerations\",\n \"mo\": 2,\n \"mn\": \"Single revision with future date suggests early development stage, no evidence of working group adoption or broader community engagement\",\n \"r\": 5,\n \"rn\": \"Directly addresses core AI agent authentication challenges, specifically targeting LLM-powered systems and their unique security vulnerabilities\",\n \"c\": [\"Agent identity/auth\", \"AI safety/alignment\", \"Human-agent interaction\"]\n}", "in_tok": 636, "out_tok": 293} {"draft_name": "draft-jadoon-nmrg-agentic-ai-autonomous-networks", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft proposes architectural principles for integrating AI agents as first-class entities at each layer of the IP protocol stack while maintaining deterministic protocol layering for interoperability. The approach introduces agent controllers to coordinate AI agents across layers using agentic methods, aiming to evolve from static automation toward goal-driven closed-loop network operations.\",\n \"n\": 4,\n \"nn\": \"Notable originality in proposing systematic agentic augmentation of existing protocol layers rather than replacing them, with coordinated multi-layer AI agent architecture\",\n \"m\": 2,\n \"mn\": \"Early conceptual sketch that introduces principles and research challenges but lacks concrete protocols, mechanisms, or implementation details\",\n \"o\": 2,\n \"on\": \"Minor similarities to general autonomous networking concepts but unique in its specific agentic augmentation approach to protocol stack layers\",\n \"mo\": 2,\n \"mn\": \"Single revision draft from research group, appears to be initiating discussion rather than building on established community work\",\n \"r\": 5,\n \"rn\": \"Core AI agent topic directly addressing autonomous network operations with agentic AI systems including planning, reasoning, and tool invocation\",\n \"c\": [\"Autonomous netops\", \"A2A protocols\", \"Policy/governance\"]\n}", "in_tok": 722, "out_tok": 297} {"draft_name": "draft-calabria-bmwg-ai-fabric-terminology", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft defines standardized benchmarking terminology for Ethernet-based network fabrics used in distributed AI training and inference workloads. It consolidates and extends existing RFC terminology with AI-specific definitions for collective communication primitives, RDMA transports, and performance metrics.\",\n \"n\": 2,\n \"nn\": \"Incremental work extending existing benchmarking terminology to AI fabric context, useful but not groundbreaking\",\n \"m\": 4,\n \"mn\": \"Well-structured terminology document with detailed definitions, examples, and clear scope as companion to methodology drafts\",\n \"o\": 3,\n \"on\": \"Explicitly builds on RFC 1242 and RFC 8238, shares concepts with networking benchmarking work but AI fabric focus is distinct\",\n \"mo\": 3,\n \"mon\": \"Active development as part of coordinated draft series, but limited evidence of broader WG momentum\",\n \"r\": 4,\n \"rn\": \"Directly relevant to AI infrastructure with focus on distributed training/inference networking, essential foundation work\",\n \"c\": [\"ML traffic mgmt\", \"Model serving/inference\"]\n}", "in_tok": 682, "out_tok": 268} {"draft_name": "draft-xiong-rtgwg-requirements-hp-wan", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft outlines requirements for high-performance wide area networks to support big data and intelligent computing applications that require massive data transmission between data centers. The focus is on ensuring data integrity and providing stable, efficient transmission services in WANs and MANs.\",\"n\":2,\"nn\":\"Incremental work identifying network requirements for data-intensive applications - useful but not particularly novel\",\"m\":2,\"mn\":\"Early requirements document without detailed technical mechanisms or protocols defined\",\"o\":3,\"on\":\"Shares networking optimization concepts with existing WAN and data center interconnect work\",\"mo\":2,\"mon\":\"Single revision with basic requirements gathering, no clear working group traction yet\",\"r\":2,\"rn\":\"Tangentially related to AI through mention of intelligent computing applications, but primarily focused on general network performance requirements\",\"c\":[\"ML traffic mgmt\"]}", "in_tok": 518, "out_tok": 182} {"draft_name": "draft-thomson-aipref-sup", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft defines a simple format for content creators to express preferences about how automated systems should consume their content. It updates RFC 9309 and defines an HTTP header field for conveying these usage preference strings.\",\"n\":2,\"nn\":\"Incremental work building on existing preference mechanisms, but addresses emerging AI content consumption concerns\",\"m\":4,\"mn\":\"Detailed specification with clear syntax definitions, HTTP header field specification, and practical examples\",\"o\":3,\"on\":\"Shares conceptual space with existing preference frameworks like RFC 7240, but focuses specifically on automated content consumption\",\"mo\":2,\"mon\":\"Single revision with no clear working group adoption or broader community engagement yet\",\"r\":4,\"rn\":\"Directly addresses the timely issue of how content creators can signal preferences to AI systems and automated processors\",\"c\":[\"Policy/governance\",\"Data formats/interop\",\"Human-agent interaction\"]}", "in_tok": 513, "out_tok": 192} {"draft_name": "draft-serra-mcp-discovery-uri", "model": "claude-sonnet-4-20250514", "raw": "```json\n{\n \"s\": \"Defines the 'mcp' URI scheme and discovery mechanisms for AI agents to find MCP servers via well-known URIs and DNS TXT records. Includes manifest integrity using JWS and security capability negotiation for trust establishment.\",\n \"n\": 3,\n \"nn\": \"Useful contribution applying standard web discovery patterns to MCP - not groundbreaking but addresses real need\",\n \"m\": 4,\n \"mn\": \"Detailed specification with clear URI scheme definition, dual discovery modes, JWS integrity, and security negotiation mechanisms\",\n \"o\": 2,\n \"on\": \"Minor similarities to general service discovery drafts but MCP-specific application is unique\",\n \"mo\": 2,\n \"mon\": \"Single revision with future date suggests early stage, unclear community adoption or working group interest\",\n \"r\": 5,\n \"rn\": \"Core AI agent topic - directly addresses agent-to-service discovery for Model Context Protocol integration\",\n \"c\": [\"Agent discovery/reg\", \"Agent identity/auth\", \"A2A protocols\"]\n}\n```", "in_tok": 699, "out_tok": 259} {"draft_name": "draft-ietf-aipref-vocab", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft defines a vocabulary for expressing preferences about how digital assets should be used by automated processing systems, allowing content creators to declare restrictions or permissions for AI systems. It addresses the growing need for machine-readable policies governing AI training and usage of copyrighted or sensitive content.\",\"n\":3,\"nn\":\"Useful contribution addressing current AI governance needs, though vocabulary-based approaches for usage rights are somewhat established in other domains\",\"m\":4,\"mn\":\"Appears to be a detailed specification with defined vocabulary terms and likely includes examples of usage patterns\",\"o\":3,\"on\":\"Shares conceptual ground with existing rights expression languages and content licensing frameworks, but focuses specifically on AI usage scenarios\",\"mo\":4,\"mon\":\"Strong working group adoption indicated by draft-ietf prefix and addresses urgent industry need for AI content usage policies\",\"r\":5,\"rn\":\"Core AI governance topic addressing fundamental questions about training data rights and automated system permissions\",\"c\":[\"Policy/governance\",\"Data formats/interop\"]}", "in_tok": 492, "out_tok": 212} {"draft_name": "draft-chen-secure-routing-use-cases", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft proposes incorporating security factors into routing decisions beyond traditional shortest-path metrics. The goal is to enable secure routing paths by considering the trustworthiness of links and forwarding nodes during route computation.\",\"n\":2,\"nn\":\"Incremental concept building on existing secure routing research, though application to modern networks has some merit\",\"m\":1,\"mn\":\"Only presents problem statement and high-level use cases without defining actual mechanisms or protocols\",\"o\":4,\"on\":\"Significant overlap with decades of secure routing research including BGPsec, S-BGP, and trust-aware routing proposals\",\"mo\":2,\"mon\":\"Appears to be single revision with basic problem statement, no evident working group interest or community engagement\",\"r\":1,\"rn\":\"This is about network routing security, not AI or autonomous agents - completely outside the AI/agent domain\",\"c\":[\"Other AI/agent\"]}", "in_tok": 615, "out_tok": 194} {"draft_name": "draft-miller-ssh-agent", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft describes the SSH Agent Protocol for securely managing SSH private keys through a key agent service. It defines message formats and procedures for key operations like signing, key listing, and key management in SSH authentication workflows.\",\"n\":1,\"nn\":\"This is documentation of an existing, well-established protocol that has been in use for decades, not a novel contribution\",\"m\":4,\"mn\":\"Comprehensive specification with detailed message formats, operation procedures, and implementation guidance\",\"o\":1,\"on\":\"Unique documentation of SSH-specific key agent protocol with no overlap with other IETF drafts\",\"mo\":2,\"mon\":\"Single revision with no apparent working group activity or community engagement\",\"r\":1,\"rn\":\"This is about SSH key management agents, not AI agents - completely unrelated to artificial intelligence\",\"c\":[\"Other AI/agent\"]}", "in_tok": 457, "out_tok": 182} {"draft_name": "draft-li-dmsc-macp", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"Proposes MACP protocol suite for multi-agent collaboration with Agent Management Centers, gateways, and capability-based discovery mechanisms. Defines architecture for secure agent onboarding, distributed synchronization, and cross-domain collaboration.\",\n \"n\": 3,\n \"nn\": \"Useful contribution addressing multi-agent coordination challenges, though builds on established distributed systems and service discovery concepts\",\n \"m\": 3,\n \"mn\": \"Defines protocol components and architectural entities with clear roles, but lacks detailed message formats and implementation specifics\",\n \"o\": 3,\n \"on\": \"Shares concepts with service discovery protocols and distributed coordination mechanisms, but applies them specifically to agent collaboration\",\n \"mo\": 2,\n \"mon\": \"Single revision with future date suggests early stage work without established community engagement\",\n \"r\": 5,\n \"rn\": \"Core AI agent topic directly addressing multi-agent system coordination, discovery, and secure collaboration protocols\",\n \"c\": [\"A2A protocols\", \"Agent discovery/reg\", \"Agent identity/auth\"]\n}", "in_tok": 574, "out_tok": 250} {"draft_name": "draft-silver-aipref-vocab-substitutive", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft proposes a vocabulary category for 'AI Substitutive Use' that allows parties to express preferences about how their digital assets are used by AI systems, specifically focusing on inference-time uses that might generate outputs substituting for original content. It's designed to nest within a broader Automated Processing category from existing work.\",\"n\":2,\"nn\":\"incremental extension of existing AI preference vocabulary work with specific focus on substitutive use cases\",\"m\":2,\"mn\":\"early sketch phase with basic category definition but lacks detailed specification or implementation guidance\",\"o\":4,\"on\":\"significant overlap with referenced AIPREF-VOCAB draft and builds directly on automated processing concepts\",\"mo\":3,\"mon\":\"active development as evidenced by recent March 2026 date and references to July 2025 working group draft\",\"r\":4,\"rn\":\"directly relevant to AI governance and policy expression for content usage by AI systems\",\"c\":[\"Policy/governance\",\"Data formats/interop\"]}", "in_tok": 576, "out_tok": 213} {"draft_name": "draft-zhang-dmsc-mas-communication", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft analyzes security limitations in current Internet protocols (TLS, HTTP, MQTT) when applied to multi-agent systems, focusing on agent-specific requirements like dynamic identity and intermediary trust. It serves as a gap analysis to inform future IETF standardization for agent-native communication security.\",\n \"n\": 3,\n \"nn\": \"Useful contribution applying established security analysis to the emerging domain of multi-agent systems, identifying specific gaps in current protocols for agent semantics\",\n \"m\": 2,\n \"mn\": \"Early-stage analysis document that identifies problems and gaps but doesn't propose concrete solutions or protocols\",\n \"o\": 2,\n \"on\": \"While security analysis is common, the specific focus on multi-agent system communication patterns and agent-native semantics has minimal overlap with existing work\",\n \"mo\": 2,\n \"mon\": \"Single revision with recent publication date but no indication of working group adoption or broader community engagement yet\",\n \"r\": 5,\n \"rn\": \"Directly addresses core AI agent communication security, covering fundamental protocols and frameworks used in multi-agent deployments\",\n \"c\": [\"A2A protocols\", \"Agent identity/auth\"]\n}", "in_tok": 601, "out_tok": 280} {"draft_name": "draft-sun-zhang-iaip", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft specifies IAIP, a protocol for AI agent interconnection at gateway level using semantic intent rather than static addressing. It defines mechanisms for agent registration, capability advertisement, intent resolution and routing decisions to enable dynamic agent discovery and task dispatching.\",\"n\":4,\"nn\":\"Intent-based semantic routing for AI agents represents a notable departure from traditional network addressing paradigms\",\"m\":3,\"mn\":\"Protocol mechanisms are defined with clear functional components but lacks detailed message formats and implementation specifics\",\"o\":2,\"on\":\"While agent discovery exists elsewhere, the intent-based semantic routing approach at gateway level appears relatively unique\",\"mo\":2,\"mon\":\"Single draft submission with no indication of working group activity or broader community engagement\",\"r\":5,\"rn\":\"Directly addresses core AI agent networking challenges with agent-to-agent communication protocol design\",\"c\":[\"A2A protocols\",\"Agent discovery/reg\"]}", "in_tok": 595, "out_tok": 193} {"draft_name": "draft-nourse-scep", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft specifies SCEP (Simple Certificate Enrollment Protocol), a PKI communication protocol using PKCS#7 and PKCS#10 over HTTP for certificate enrollment. SCEP was originally developed by VeriSign for Cisco and now has wide implementation support across client and CA systems.\",\"n\":1,\"nn\":\"This is standardization of an existing, well-established protocol rather than novel work\",\"m\":4,\"mn\":\"Appears to be a detailed specification of a mature protocol with 43 pages of content\",\"o\":4,\"on\":\"SCEP is a well-known existing protocol with multiple existing implementations and likely previous standardization efforts\",\"mo\":2,\"mon\":\"Single draft submission with no clear indication of working group adoption or broad IETF community engagement\",\"r\":1,\"rn\":\"This is a traditional PKI certificate enrollment protocol with no AI or agent-specific components\",\"c\":[\"Other AI/agent\"]}", "in_tok": 537, "out_tok": 206} {"draft_name": "draft-mp-agntcy-ads", "model": "claude-sonnet-4-20250514", "raw": "```json\n{\n \"s\": \"The Agent Directory Service (ADS) is a distributed directory service for storing and discovering AI agent metadata based on skills and capabilities. It uses interconnected distributed directories with a content-routing protocol to enable agent discovery for problem-solving applications.\",\n \"n\": 3,\n \"nn\": \"Applies well-established distributed directory concepts to AI agent discovery, which is a useful contribution to emerging agent ecosystems\",\n \"m\": 4,\n \"mn\": \"At 31 pages, likely contains detailed protocol specifications, data structures, and operational mechanisms for the distributed directory service\",\n \"o\": 2,\n \"on\": \"While directory services exist, applying them specifically to AI agent skill-based discovery appears to have minimal overlap with existing IETF work\",\n \"mo\": 2,\n \"mon\": \"Single draft submission without clear working group adoption or broader community engagement visible\",\n \"r\": 5,\n \"rn\": \"Directly addresses core AI agent infrastructure needs for discovery and registration in distributed agent systems\",\n \"c\": [\"Agent discovery/reg\", \"A2A protocols\", \"Data formats/interop\"]\n}\n```", "in_tok": 509, "out_tok": 267} {"draft_name": "draft-sato-soos-gar", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft specifies the Governance Audit Record (GAR) system for auditing agentic AI systems, defining five audit types, alert systems, and audit packages for regulatory inspection. GAR provides verifiable evidence that AI agent sessions comply with governance frameworks and can be proven to regulators.\",\n \"n\": 4,\n \"nn\": \"Novel approach to AI governance auditing with comprehensive audit architecture including multiple record types and regulatory compliance focus\",\n \"m\": 3,\n \"mn\": \"Defines clear audit architecture and record types but likely needs more implementation details and examples for full specification maturity\",\n \"o\": 2,\n \"on\": \"References related drafts in same series (IDP, HEM) but appears to address unique audit and governance verification aspects\",\n \"mo\": 3,\n \"mon\": \"Part of active draft series with consistent authorship and forward-looking 2026 date suggesting ongoing development\",\n \"r\": 5,\n \"rn\": \"Directly addresses core AI agent governance and regulatory compliance challenges, essential for enterprise AI deployment\",\n \"c\": [\"Policy/governance\", \"Human-agent interaction\"]\n}", "in_tok": 606, "out_tok": 267} {"draft_name": "draft-mw-spice-inference-chain", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"Defines cryptographic inference chains to prove AI models actually performed claimed computations, using ZKML proofs and TEE attestation. Part of a three-chain 'Truth Stack' covering delegation (WHO), content (WHAT), and computational (HOW) provenance for AI agents.\",\n \"n\": 4,\n \"nn\": \"Novel application of ZKML and TEE attestation to AI computational provenance, creating verifiable inference chains - addresses a gap not covered by existing standards\",\n \"m\": 3,\n \"mn\": \"Defines protocol mechanisms and integrates with OAuth tokens, but lacks implementation details and test vectors for production deployment\",\n \"o\": 2,\n \"on\": \"Complements related drafts in same series (actor-chain, intent-chain) but addresses distinct computational provenance problem with unique technical approach\",\n \"mo\": 3,\n \"mon\": \"Active development as part of coordinated multi-draft effort, but uncertain broader community adoption beyond the draft series authors\",\n \"r\": 5,\n \"rn\": \"Directly addresses core AI agent governance challenge of proving computational integrity - essential for trustworthy autonomous systems\",\n \"c\": [\"Policy/governance\", \"Agent identity/auth\", \"Model serving/inference\"]\n}", "in_tok": 704, "out_tok": 294} {"draft_name": "draft-templin-intarea-aero", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"AERO specifies an automatic route optimization service for IP internetworking over overlay multilink networks, using IPv6 neighbor discovery for network admission and mobility management. The protocol targets mobile applications across air/land/sea/space domains including aviation and intelligent transportation systems.\",\n \"n\": 3,\n \"nn\": \"Builds on existing IPv6 ND and overlay networking concepts but provides useful integration for mobility scenarios\",\n \"m\": 4,\n \"mn\": \"Detailed 120-page specification with comprehensive protocol mechanisms and clear technical approach\",\n \"o\": 3,\n \"on\": \"Shares overlay networking concepts with other mobility and tunneling drafts but has distinct OMNI interface approach\",\n \"mo\": 2,\n \"mon\": \"Single author draft with limited apparent working group engagement or community discussion\",\n \"r\": 1,\n \"rn\": \"This is a networking mobility protocol with no AI or agent-related content - appears to be a false positive\",\n \"c\": [\"Other AI/agent\"]\n}", "in_tok": 656, "out_tok": 243} {"draft_name": "draft-wang-lisp-ai-agent", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft proposes using LISP (Locator/ID Separation Protocol) as a network substrate for AI agent communication, addressing requirements like location transparency and seamless mobility. It outlines minimal extensions to LISP for context-aware mapping and AI agent-centric communication patterns.\",\"n\":3,\"nn\":\"Useful application of existing LISP to AI agents, though the core insight of using ID/locator separation for agent mobility is somewhat predictable\",\"m\":2,\"mn\":\"Early stage work with basic framework outlined but lacking detailed protocol specifications or implementation guidance\",\"o\":2,\"on\":\"While LISP is well-established, its specific application to AI agent networking has limited overlap with existing drafts\",\"mo\":2,\"mon\":\"Single revision from 2026 with no clear indication of working group interest or community uptake\",\"r\":4,\"rn\":\"Directly addresses AI agent networking requirements with concrete technical approach, highly relevant to distributed AI systems\",\"c\":[\"A2A protocols\",\"Agent discovery/reg\",\"Autonomous netops\"]}", "in_tok": 586, "out_tok": 227} {"draft_name": "draft-mw-spice-actor-chain", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft defines six cryptographically verifiable actor-chain profiles for OAuth 2.0 Token Exchange to preserve and validate delegation paths across multi-hop service-to-service and agentic workflows. It extends RFC 8693 with standardized mechanisms for tracking actor chains while providing different privacy and auditability tradeoffs.\",\n \"n\": 4,\n \"nn\": \"Notable contribution addressing a clear gap in OAuth 2.0 Token Exchange for multi-hop agentic workflows with novel cryptographic verification profiles\",\n \"m\": 4,\n \"mn\": \"Detailed 97-page specification with six well-defined profiles, specific processing rules, and comprehensive coverage of delegation scenarios\",\n \"o\": 2,\n \"on\": \"Builds on RFC 8693 but addresses previously unspecified delegation-path validation with minimal overlap to existing OAuth extensions\",\n \"mo\": 3,\n \"mon\": \"Single draft revision but addresses emerging need for agentic workflow authentication with reasonable technical approach\",\n \"r\": 4,\n \"rn\": \"Directly relevant to AI agent authentication and authorization, specifically targeting agentic workflows and service-to-service delegation chains\",\n \"c\": [\"Agent identity/auth\", \"A2A protocols\"]\n}", "in_tok": 734, "out_tok": 294} {"draft_name": "draft-pocero-authkem-edhoc", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft extends the EDHOC protocol to provide post-quantum cryptography resistance by incorporating KEM-based authentication mechanisms. It defines signature-free authentication using NIST-standardized ML-KEM and supports hybrid scenarios mixing KEM-based and PQC signature authentication.\",\"n\":3,\"nn\":\"Useful contribution applying established PQC KEMs to EDHOC authentication, though building on known cryptographic primitives\",\"m\":4,\"mn\":\"Detailed specification with concrete protocol extensions, KEM integration mechanisms, and hybrid authentication scenarios\",\"o\":2,\"on\":\"Minor overlap with general PQC protocol work but unique in its specific KEM-based authentication approach for EDHOC\",\"mo\":3,\"mon\":\"Active development indicated by 2026 date and substantial 42-page length, though momentum unclear without revision history\",\"r\":2,\"rn\":\"Tangentially related to AI/agents only through general authentication mechanisms that could be used by agent systems\",\"c\":[\"Agent identity/auth\"]}", "in_tok": 624, "out_tok": 221} {"draft_name": "draft-c4tz-marc", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"MARC defines a vendor-neutral profile for generative models and agents to disclose uncertainty and control metadata, separating capability assessment from answer confidence. It specifies a bounded action set including answering, clarification, retrieval, tool use, deliberation, abstention, and escalation to improve auditability and reduce silent failures.\",\n \"n\": 4,\n \"nn\": \"Notable originality in separating pre-decision capability assessment from post-decision confidence and defining a bounded primary action set for agent behaviors\",\n \"m\": 3,\n \"mn\": \"Defined protocol with clear separation of concerns and bounded action semantics, though lacking detailed implementation examples\",\n \"o\": 2,\n \"on\": \"Minor similarities to general API standards but unique approach to uncertainty disclosure and agent control semantics\",\n \"mo\": 2,\n \"mon\": \"Single revision with future date suggests early stage development without demonstrated community engagement\",\n \"r\": 5,\n \"rn\": \"Core AI agent topic addressing fundamental issues of uncertainty communication, control, and interoperability in generative systems\",\n \"c\": [\"AI safety/alignment\", \"Data formats/interop\", \"Human-agent interaction\", \"Model serving/inference\"]\n}", "in_tok": 677, "out_tok": 283} {"draft_name": "draft-tejido-swp-core", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"SlimWire Protocol (SWP) defines a binary framing layer and envelope format for agent-to-agent messaging with profile-based extensibility. The spec includes conformance testing via golden vectors and requires authenticated confidential channels for security.\",\n \"n\": 3,\n \"nn\": \"Useful binary messaging protocol specifically designed for agent communication with extensible profiles, though builds on established framing concepts\",\n \"m\": 4,\n \"mn\": \"Detailed specification with defined envelope encoding (E1), security binding (S1), conformance classes, and golden vector test artifacts\",\n \"o\": 2,\n \"on\": \"While binary framing exists elsewhere, the agent-specific profile dispatch and conformance model represent a relatively unique approach\",\n \"mo\": 2,\n \"mon\": \"Single draft from 2026 with no visible working group adoption or community discussion yet\",\n \"r\": 5,\n \"rn\": \"Core infrastructure protocol specifically designed for agent-to-agent communication with tool invocation mapping and agent task lifecycle support\",\n \"c\": [\"A2A protocols\", \"Data formats/interop\"]\n}", "in_tok": 681, "out_tok": 266} {"draft_name": "draft-rosenberg-aiproto-framework", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft proposes a framework for AI agent communications on the Internet, identifying protocols needed for agent-to-agent and agent-to-resource interactions. It surveys existing work like MCP and A2A protocols while establishing requirements for potential IETF standardization in this emerging space.\",\n \"n\": 3,\n \"nn\": \"Provides a useful framework for organizing AI agent protocol requirements, though the core concepts of agent communication are becoming well-established in the field\",\n \"m\": 2,\n \"mn\": \"Early framework document that identifies problem space and surveys existing work but lacks concrete protocol definitions or detailed specifications\",\n \"o\": 3,\n \"on\": \"Shares conceptual overlap with MCP and A2A protocol work mentioned in the abstract, plus general agent communication frameworks in the broader AI community\",\n \"mo\": 3,\n \"mon\": \"Shows active development as a 2026 draft addressing timely AI agent standardization needs, though IETF momentum in this space is still building\",\n \"r\": 5,\n \"rn\": \"Core AI agent topic directly addressing fundamental protocols needed for agent communication, interoperability, and Internet-scale deployment\",\n \"c\": [\"A2A protocols\", \"Agent discovery/reg\", \"Data formats/interop\", \"Human-agent interaction\"]\n}", "in_tok": 699, "out_tok": 305} {"draft_name": "draft-mandel-lamps-rfc5274bis", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft updates compliance requirements for the Certificate Management over CMS (CMC) enrollment protocol, obsoleting RFCs 5274 and 6402. It provides implementation guidance for creating compliant CMC systems without defining new ASN.1 structures or transport mechanisms.\",\"n\":1,\"nn\":\"Trivial update to existing certificate management compliance requirements with no novel technical contributions\",\"m\":4,\"mn\":\"Detailed compliance specification that appears implementation-ready, building on well-established CMC protocol foundations\",\"o\":5,\"on\":\"Direct successor document that obsoletes two existing RFCs covering the same compliance requirements for CMC\",\"mo\":3,\"mon\":\"Standard maintenance update showing active development but limited broader community engagement signals\",\"r\":2,\"rn\":\"Certificate management has tangential relevance to agent authentication but this compliance document doesn't address AI/agent-specific concerns\",\"c\":[\"Agent identity/auth\"]}", "in_tok": 540, "out_tok": 197} {"draft_name": "draft-jia-oauth-scope-aggregation", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"Proposes OAuth 2.0 scope aggregation to reduce authorization friction in multi-step AI agent workflows by collecting all required permissions upfront. Aims to minimize repeated user consent prompts when agents need to access multiple resources across a workflow.\",\n \"n\": 2,\n \"nn\": \"Incremental extension of OAuth 2.0 for AI workflows - addresses a real problem but solution is relatively straightforward application of existing patterns\",\n \"m\": 3,\n \"mn\": \"Defines a clear protocol mechanism for scope aggregation with workflow-aware authorization, though likely needs more implementation details\",\n \"o\": 3,\n \"on\": \"Shares concepts with OAuth extensions and AI agent authorization work, but specific focus on workflow scope aggregation is reasonably distinct\",\n \"mo\": 2,\n \"mon\": \"February 2026 draft suggests early stage work with limited revision history or demonstrated community engagement\",\n \"r\": 4,\n \"rn\": \"Directly addresses authorization challenges for AI agents that execute multi-step workflows requiring diverse resource access permissions\",\n \"c\": [\"Agent identity/auth\", \"Human-agent interaction\"]\n}", "in_tok": 535, "out_tok": 268} {"draft_name": "draft-ietf-anima-constrained-voucher", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft defines cBRSKI, a constrained version of BRSKI for secure zero-touch onboarding of IoT devices using CBOR-encoded vouchers instead of JSON and CoAPS instead of HTTPS. It adapts the existing BRSKI protocol for resource-constrained networks with limited throughput and frequent packet loss.\",\"n\":2,\"nn\":\"Incremental adaptation of existing BRSKI protocol for constrained environments, primarily involving format changes (JSON to CBOR) and transport substitutions (HTTPS to CoAPS)\",\"m\":4,\"mn\":\"Detailed specification with comprehensive protocol definition, clear examples, and implementation guidance across 93 pages\",\"o\":4,\"on\":\"Significant overlap with RFC 8995 (BRSKI) and RFC 9148, as this is explicitly a constrained variant that reuses core concepts and mechanisms\",\"mo\":4,\"mon\":\"Strong working group adoption as an IETF draft with clear update relationship to established RFCs, indicating active development and community support\",\"r\":2,\"rn\":\"Tangentially related to AI/agents - focuses on IoT device bootstrapping and authentication rather than AI agent systems, though secure device onboarding could support AI-enabled IoT deployments\",\"c\":[\"Agent identity/auth\"]}", "in_tok": 735, "out_tok": 284} {"draft_name": "draft-du-ai-agent-communication-6g-aspect", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft proposes requirements and use cases for AI Agent Communication Networks (ACN) in 6G systems, focusing on agent-to-agent communication as a new traffic type. It discusses 3GPP use cases and potential frameworks for enabling AI agents to communicate and collaborate across networks.\",\"n\":2,\"nn\":\"Incremental work applying existing networking concepts to AI agent communication without significant technical innovation\",\"m\":2,\"mn\":\"Early-stage document with high-level requirements and use cases but lacks concrete protocol definitions or technical mechanisms\",\"o\":3,\"on\":\"Shares concepts with other 6G and AI agent networking proposals, moderate overlap with existing mobile network research\",\"mo\":3,\"mon\":\"Active development indicated by version 6, appears to have ongoing work but unclear community adoption\",\"r\":4,\"rn\":\"Directly addresses AI agent networking requirements with specific focus on agent-to-agent communication protocols\",\"c\":[\"A2A protocols\",\"ML traffic mgmt\",\"Autonomous netops\",\"Agent discovery/reg\"]}", "in_tok": 610, "out_tok": 218} {"draft_name": "draft-tschofenig-lamps-nonce-cmp-est", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft specifies how nonces are provided by Registration/Certification Authorities to end entities for inclusion in attestation evidence within Certificate Signing Requests for CMP and EST protocols. It addresses freshness requirements for remote attestation in PKI certificate management scenarios.\",\n \"n\": 2,\n \"nn\": \"Incremental extension applying existing nonce-based freshness concepts to CMP/EST certificate management protocols\",\n \"m\": 3,\n \"mn\": \"Defines specific protocol mechanisms for nonce handling in CSRs with clear technical approach\",\n \"o\": 3,\n \"on\": \"Shares concepts with existing attestation and PKI management specifications but addresses specific gap\",\n \"mo\": 2,\n \"mon\": \"Single revision draft with limited visible community engagement or working group adoption\",\n \"r\": 2,\n \"rn\": \"Tangentially related through attestation mechanisms but primarily focused on PKI certificate management rather than AI agents\",\n \"c\": [\"Agent identity/auth\"]\n}", "in_tok": 670, "out_tok": 239} {"draft_name": "draft-illyes-aipref-cbcp", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft proposes best practices for web crawlers, focusing on responsible crawling behavior and server interaction guidelines. While crawler behavior is relevant to AI agents that scrape web content, this appears to be a general web crawling guidance document rather than AI-specific protocol work.\",\"n\":2,\"nn\":\"incremental contribution to existing crawler guidance\",\"m\":2,\"mn\":\"early sketch with limited technical detail provided\",\"o\":4,\"on\":\"significant overlap with existing robots.txt and crawler standards\",\"mo\":2,\"mon\":\"appears to be single author draft with limited activity\",\"r\":2,\"rn\":\"tangentially related - crawlers used by AI but not AI-specific\",\"c\":[\"Policy/governance\"]}", "in_tok": 500, "out_tok": 153} {"draft_name": "draft-xu-efficient-agent-discovery-profile", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft defines a standardized metadata and query profile for discovering autonomous software agents, specifying JSON objects for publishing agent capabilities and enabling interoperable discovery services. The profile allows structured filtering and semantic matching while remaining agnostic to specific ML models or databases.\",\"n\":3,\"nn\":\"Useful contribution addressing a clear need for agent discovery standardization, though the approach of metadata profiles for service discovery follows established patterns\",\"m\":4,\"mn\":\"Detailed specification with comprehensive JSON schemas, examples, and clear protocol definitions that appear implementation-ready\",\"o\":2,\"on\":\"While service discovery is well-established, the specific focus on autonomous agent capabilities and multi-modal matching criteria has minor similarities to existing work\",\"mo\":2,\"mon\":\"Single revision with 2026 date suggests early stage without visible working group adoption or community engagement\",\"r\":5,\"rn\":\"Directly addresses core AI agent infrastructure needs for discovery and interoperability in multi-agent systems\",\"c\":[\"Agent discovery/reg\",\"Data formats/interop\",\"A2A protocols\"]}", "in_tok": 622, "out_tok": 224} {"draft_name": "draft-liu-saag-zt-problem-statement", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft presents a problem statement for zero trust standardization in IETF, arguing that while high-level architectural guidance exists (like NIST SP 800-207), the industry lacks interoperable frameworks and protocols for multi-vendor zero trust implementations. The document outlines use cases including cloud services, remote workforces, and intelligent agents, proposing IETF standardization work to enable zero trust interoperability.\",\n \"n\": 2,\n \"nn\": \"Incremental contribution applying well-established zero trust concepts to IETF standardization context, with brief mention of intelligent agents but no novel technical insights\",\n \"m\": 1,\n \"mn\": \"Pure problem statement document with numerous typos and grammatical errors, lacks any concrete protocol definitions or technical mechanisms\",\n \"o\": 4,\n \"on\": \"Significant overlap with existing zero trust architecture work (NIST SP 800-207) and various IETF security protocols, though focuses on interoperability gap\",\n \"mo\": 2,\n \"mon\": \"Single revision with future date (2026-05-06), no evidence of working group interest or community engagement\",\n \"r\": 2,\n \"rn\": \"Only tangentially related to AI/agents through brief mention of 'intelligent agents' as a use case, primarily focused on general zero trust networking\",\n \"c\": [\"Policy/governance\", \"Agent identity/auth\"]\n}", "in_tok": 736, "out_tok": 332} {"draft_name": "draft-kwiatkowski-tls-ecdhe-mlkem", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"Defines hybrid post-quantum key agreement methods for TLS 1.3 combining MLKEM with classical ECDHE. Specifies three concrete combinations: X25519MLKEM768, SecP256r1MLKEM768, and SecP384r1MLKEM1024.\",\"n\":2,\"nn\":\"Incremental work combining established post-quantum and classical cryptographic primitives in straightforward manner\",\"m\":4,\"mn\":\"Detailed specification with concrete parameter choices, wire formats, and implementation guidance\",\"o\":4,\"on\":\"Significant overlap with other post-quantum TLS hybrid proposals and existing NIST standardization efforts\",\"mo\":4,\"mon\":\"Strong industry momentum behind post-quantum cryptography adoption with active IETF working group development\",\"r\":1,\"rn\":\"Pure cryptographic protocol specification with no AI or agent-related functionality\",\"c\":[\"Other AI/agent\"]}", "in_tok": 542, "out_tok": 202} {"draft_name": "draft-hood-independent-agis", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"AGIS defines a grammar-constrained interface language for LLM-based agents to interact with APIs using intent-expressing imperative verbs as method names. It serves as the native interface layer for the Agent Transfer Protocol (AGTP), similar to how HTML functions for HTTP.\",\n \"n\": 4,\n \"nn\": \"Notable originality in defining grammar constraints for agent-API interactions with empirical validation showing 10-29 percentage point accuracy improvements over generic HTTP verbs\",\n \"m\": 4,\n \"mn\": \"Detailed 50-page specification with YAML/JSON serialization formats, data manifest blocks, semantic declarations, and HTTP transport mapping\",\n \"o\": 2,\n \"on\": \"Minor similarities to existing API definition languages but unique focus on LLM agent consumption and grammar-constrained method naming\",\n \"mo\": 2,\n \"mon\": \"Single draft with no clear working group adoption or broader community engagement visible\",\n \"r\": 5,\n \"rn\": \"Core AI agent topic addressing fundamental interface definition needs for LLM-based agents with empirical validation\",\n \"c\": [\"A2A protocols\", \"Data formats/interop\"]\n}", "in_tok": 899, "out_tok": 279} {"draft_name": "draft-illyes-aipref-jafar", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft defines a standardized JSON format for automated HTTP clients (crawlers, AI bots) to publicly disclose their IP address ranges. The format aims to help website operators identify legitimate automated traffic and reduce inadvertent blocking of beneficial clients.\",\"n\":2,\"nn\":\"Incremental improvement on existing ad-hoc practices for IP range disclosure, providing standardization but no fundamental innovation\",\"m\":4,\"mn\":\"Well-defined specification with clear JSON schema, examples, and implementation guidance ready for deployment\",\"o\":3,\"on\":\"Shares concepts with existing network management and traffic identification approaches, but specific to automated clients\",\"mo\":2,\"mon\":\"Single revision draft with no clear working group adoption or community engagement signals\",\"r\":4,\"rn\":\"Directly addresses AI bot and automated agent traffic management, a growing operational concern for web infrastructure\",\"c\":[\"ML traffic mgmt\",\"Agent identity/auth\",\"Data formats/interop\"]}", "in_tok": 577, "out_tok": 201} {"draft_name": "draft-sogomonian-ai-uri-scheme", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft proposes an experimental 'ai:' URI scheme to provide dedicated access points for AI resources, enabling native connectivity for autonomous systems while supporting human-facing applications through HTTPS gateways. The scheme aims to create a standardized way for AI systems and robots to discover and connect to AI resources across networks.\",\"n\":3,\"nn\":\"Useful contribution creating dedicated URI namespace for AI resources, though URI schemes for specialized domains have precedent\",\"m\":2,\"mn\":\"Early sketch with basic scheme definition but lacks detailed protocol mechanics, registration procedures, and security considerations\",\"o\":2,\"on\":\"Minor similarities to existing URI schemes but AI-specific application is relatively unique\",\"mo\":2,\"mon\":\"Single revision with future date suggests early stage without demonstrated community engagement\",\"r\":4,\"rn\":\"Directly relevant to AI agent infrastructure, addressing fundamental discovery and addressing needs for autonomous systems\",\"c\":[\"A2A protocols\",\"Agent discovery/reg\",\"Data formats/interop\"]}", "in_tok": 509, "out_tok": 209} {"draft_name": "draft-zhao-nmrg-ai-agent-for-ndt", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft proposes an AI agent architecture for Network Digital Twin (NDT) systems, integrating AI agents with digital twin technology for network management. The document appears to be an early conceptual framework for applying AI agents to digital representations of network infrastructure.\",\"n\":3,\"nn\":\"Useful contribution combining AI agents with network digital twins, though both concepts exist separately - the integration approach shows practical value\",\"m\":2,\"mn\":\"Early sketch stage with basic architecture outline, lacks detailed protocol specifications, implementation details, and concrete examples\",\"o\":3,\"on\":\"Shares concepts with existing network management and digital twin drafts, but the specific AI agent integration angle provides some differentiation\",\"mo\":2,\"mon\":\"Single revision draft with no clear working group adoption or community feedback visible yet\",\"r\":4,\"rn\":\"Directly relevant to AI agents in networking context, addressing autonomous network operations through digital twin technology\",\"c\":[\"Autonomous netops\",\"Other AI/agent\"]}", "in_tok": 477, "out_tok": 206} {"draft_name": "draft-ietf-lake-authz", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft specifies ELA (Lightweight Authorization using EDHOC), a protocol for authorizing enrollment of new devices in constrained networks using the EDHOC key exchange protocol. It enables zero-touch onboarding of devices leveraging trust anchors installed during manufacturing.\",\"n\":2,\"nn\":\"Incremental extension of existing EDHOC protocol for device authorization - useful but straightforward application\",\"m\":4,\"mn\":\"Detailed 44-page specification with clear protocol definitions and procedures for device enrollment\",\"o\":3,\"on\":\"Shares device onboarding and authorization concepts with other IoT security drafts but focuses specifically on EDHOC integration\",\"mo\":4,\"mon\":\"IETF working group draft with specific publication target date indicates active development and institutional support\",\"r\":2,\"rn\":\"Focuses on IoT device authorization rather than AI agents - only tangentially related through general authentication mechanisms\",\"c\":[\"Other AI/agent\"]}", "in_tok": 575, "out_tok": 209} {"draft_name": "draft-ra-emu-pqc-eapaka", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft proposes enhancing EAP-AKA' Forward Secrecy with Post-Quantum Key Encapsulation Mechanisms to protect against future quantum computer threats. It replaces the traditional ECDHE key exchange with quantum-resistant alternatives while maintaining the existing authentication framework.\",\n \"n\": 2,\n \"nn\": \"Incremental upgrade replacing ECDHE with PQ-KEMs in existing EAP-AKA' FS protocol\",\n \"m\": 2,\n \"mn\": \"Early specification stage at 15 pages, likely missing detailed protocol flows and implementation specifics\",\n \"o\": 4,\n \"on\": \"Significant conceptual overlap with general post-quantum cryptography migration efforts and other PQ-KEM integration drafts\",\n \"mo\": 3,\n \"mon\": \"Active development with recent July 2025 publication date, but uncertain working group status\",\n \"r\": 1,\n \"rn\": \"Pure cryptographic protocol enhancement with no AI or agent-related content - appears to be false positive\",\n \"c\": [\"Other AI/agent\"]\n}", "in_tok": 710, "out_tok": 267} {"draft_name": "draft-li-zt-consideration", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft examines applying Zero Trust principles to network infrastructure, arguing that traditional perimeter-based security models are inadequate for modern distributed networks. It advocates for continuous verification of all network entities and communications regardless of location to prevent lateral movement and internal compromise.\",\"n\":2,\"nn\":\"Incremental application of well-established Zero Trust concepts to network infrastructure - useful but not particularly novel\",\"m\":2,\"mn\":\"Early conceptual document that examines problems and philosophy but lacks concrete protocols, mechanisms, or implementation details\",\"o\":3,\"on\":\"Shares fundamental Zero Trust concepts with existing security frameworks and likely overlaps with other network security drafts\",\"mo\":2,\"mon\":\"Single revision with no indication of working group adoption or broader community engagement\",\"r\":1,\"rn\":\"This is a traditional network security document about Zero Trust principles with no AI or agent-related content\",\"c\":[\"Other AI/agent\"]}", "in_tok": 668, "out_tok": 195} {"draft_name": "draft-oauth-ai-agents-on-behalf-of-user", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft extends OAuth 2.0 to enable AI agents to securely obtain access tokens for acting on behalf of users through new requested_actor and actor_token parameters. It provides a framework for explicit user consent and delegation chains with enhanced auditability for AI agent authorization flows.\",\"n\":3,\"nn\":\"Useful contribution addressing AI agent delegation in OAuth, though builds incrementally on existing token delegation patterns\",\"m\":4,\"mn\":\"Detailed specification with clear protocol mechanics, parameter definitions, and flow descriptions approaching implementation readiness\",\"o\":2,\"on\":\"Some conceptual similarity to token exchange and delegation flows but introduces AI-specific mechanisms not directly covered elsewhere\",\"mo\":2,\"mon\":\"Single revision from February 2026 with no visible working group adoption or community discussion yet\",\"r\":5,\"rn\":\"Directly addresses core AI agent authentication and authorization challenges, highly relevant to AI agent ecosystem needs\",\"c\":[\"Agent identity/auth\",\"Human-agent interaction\",\"Policy/governance\"]}", "in_tok": 634, "out_tok": 212} {"draft_name": "draft-ietf-tls-extended-key-update", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft extends TLS 1.3's key update mechanism to include fresh Diffie-Hellman exchanges within active sessions, providing post-compromise security that the current KeyUpdate mechanism lacks. The extension is designed for long-lived sessions in industrial IoT and telecommunications environments where static key compromise poses significant risks.\",\n \"n\": 3,\n \"nn\": \"Useful contribution addressing a real limitation in TLS 1.3's forward secrecy guarantees for long-lived sessions\",\n \"m\": 4,\n \"mn\": \"Detailed specification with protocol mechanisms defined, appears implementation-ready\",\n \"o\": 2,\n \"on\": \"Some conceptual similarity to existing key rotation mechanisms but addresses specific TLS 1.3 gap\",\n \"mo\": 4,\n \"mon\": \"IETF working group draft with clear security need and industry applications\",\n \"r\": 1,\n \"rn\": \"This is a TLS security extension, not related to AI agents or artificial intelligence\",\n \"c\": []\n}", "in_tok": 719, "out_tok": 247} {"draft_name": "draft-irtf-nmrg-ai-deploy", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This IETF draft provides considerations for configuring networks and systems to support distributed AI inference services, covering deployment from cloud to edge devices. It addresses factors for client connections to AI services and includes use cases like self-driving vehicles and network digital twins.\",\n \"n\": 2,\n \"nn\": \"Incremental work applying existing networking concepts to AI deployment scenarios without introducing novel technical approaches\",\n \"m\": 2,\n \"mn\": \"Early sketch stage - provides high-level considerations and use cases but lacks detailed protocols, mechanisms, or implementation specifics\",\n \"o\": 4,\n \"on\": \"Significant overlap with existing edge computing, distributed systems, and AI infrastructure deployment literature and standards\",\n \"mo\": 3,\n \"mon\": \"Active development within IRTF NMRG context, though unclear level of broader community engagement beyond research group\",\n \"r\": 4,\n \"rn\": \"Directly relevant to AI agent deployment infrastructure, addressing key networking considerations for distributed AI services\",\n \"c\": [\"Model serving/inference\", \"ML traffic mgmt\", \"Autonomous netops\"]\n}", "in_tok": 611, "out_tok": 261} {"draft_name": "draft-aiendpoint-ai-discovery", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"Defines a standardized /.well-known/ai endpoint for web services to expose machine-readable capability descriptions to AI agents. Creates a structured JSON format optimized for LLM token efficiency to enable programmatic service discovery without parsing human documentation.\",\n \"n\": 3,\n \"nn\": \"Useful contribution applying well-known URI pattern to AI agent discovery - logical but not obvious extension\",\n \"m\": 4,\n \"mn\": \"Detailed specification with structured JSON schema, authentication handling, and LLM optimization considerations\",\n \"o\": 2,\n \"on\": \"Some conceptual overlap with OpenAPI/service discovery but distinct focus on AI agent consumption\",\n \"mo\": 2,\n \"mn\": \"Single revision with no clear working group adoption or community feedback yet\",\n \"r\": 5,\n \"rn\": \"Directly addresses core AI agent infrastructure need for automated service discovery and capability exposure\",\n \"c\": [\"Agent discovery/reg\", \"Data formats/interop\", \"A2A protocols\"]\n}", "in_tok": 601, "out_tok": 243} {"draft_name": "draft-abbott-mcp-ax", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"MCP-AX adapts the SNMP AgentX protocol architecture to enable hierarchical aggregation of Model Context Protocol (MCP) servers, allowing AI agents to access distributed tool namespaces across heterogeneous networks. The protocol provides recursive namespace delegation and capability-aware routing for composing tool collections from cloud services to embedded devices.\",\n \"n\": 4,\n \"nn\": \"Notable originality in adapting proven SNMP aggregation patterns to the emerging MCP ecosystem, creating a novel hierarchical tool composition architecture\",\n \"m\": 4,\n \"mn\": \"Detailed specification with clear master-subagent architecture, namespace delegation mechanisms, and transport bridging - appears implementation-ready\",\n \"o\": 2,\n \"on\": \"Some conceptual similarity to service mesh patterns but unique application to MCP tool aggregation with minimal direct overlap\",\n \"mo\": 3,\n \"mon\": \"Active development indicated by 2026 date and substantial 25-page specification, though community adoption unclear for emerging MCP ecosystem\",\n \"r\": 5,\n \"rn\": \"Core AI agent topic addressing critical infrastructure needs for tool composition and namespace management in distributed AI systems\",\n \"c\": [\"A2A protocols\", \"Agent discovery/reg\", \"Data formats/interop\"]\n}", "in_tok": 634, "out_tok": 301} {"draft_name": "draft-pang-agents-networking-scenarios", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft outlines networking scenarios for AI agents in enterprise and home broadband networks, distinct from mobile and internet deployments. The document is largely speculative, based on assumptions about emerging agentic services in these network environments.\",\"n\":2,\"nn\":\"Incremental extension of agent networking concepts to enterprise/broadband contexts without significant technical innovation\",\"m\":1,\"mn\":\"Appears to be primarily a problem statement and scenario description without concrete protocols or mechanisms\",\"o\":3,\"on\":\"Likely shares conceptual overlap with general agent networking drafts but focuses on specific deployment environments\",\"mo\":2,\"mon\":\"Single revision of an early-stage draft with no clear working group adoption signals\",\"r\":4,\"rn\":\"Directly addresses AI agent networking scenarios, though limited to specific network deployment contexts\",\"c\":[\"A2A protocols\",\"Autonomous netops\",\"Agent discovery/reg\"]}", "in_tok": 514, "out_tok": 190} {"draft_name": "draft-kiliram-agent-trust-auth-framework", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft proposes a cross-domain trust and authentication framework for AI agent-to-agent communications, addressing identity verification, authorization, and delegation across different trust domains. The framework aims to be protocol-agnostic and provide a consistent baseline for existing and emerging AI agent protocols.\",\n \"n\": 4,\n \"nn\": \"Notable originality in addressing the specific challenge of cross-domain trust for AI agents, which is emerging as protocols mature\",\n \"m\": 2,\n \"mn\": \"Early architectural framework stage - 37 pages suggests some detail but likely lacks concrete protocol specifications and implementation guidance\",\n \"o\": 2,\n \"on\": \"Some conceptual overlap with general identity/auth frameworks, but the AI agent cross-domain focus appears relatively unique\",\n \"mo\": 3,\n \"mn\": \"Active development indicated by 2026 date, but momentum unclear without working group adoption information\",\n \"r\": 5,\n \"rn\": \"Core AI agent topic directly addressing critical trust infrastructure needs for agent ecosystems\",\n \"c\": [\"Agent identity/auth\", \"A2A protocols\", \"Policy/governance\"]\n}", "in_tok": 616, "out_tok": 265} {"draft_name": "draft-peterson-mimi-idprover", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"Defines a third-party attestation mechanism for verifying telephone number ownership in messaging platforms, supporting secure enrollment in interoperable messaging systems. The approach extends to other Service Independent Identifiers in the MIMI framework for cross-platform identity verification.\",\n \"n\": 2,\n \"nn\": \"Incremental work applying standard proof-of-possession concepts to messaging platform identity verification\",\n \"m\": 3,\n \"mn\": \"Defined mechanism with clear protocol elements but lacks detailed implementation specifics and examples\",\n \"o\": 3,\n \"on\": \"Shares identity attestation concepts with existing work but addresses specific MIMI messaging interoperability context\",\n \"mo\": 3,\n \"mon\": \"Active individual draft with clear use case alignment to ongoing MIMI working group efforts\",\n \"r\": 1,\n \"rn\": \"Not about AI or agents - focuses on human user identity verification for messaging platforms\",\n \"c\": [\"Other AI/agent\"]\n}", "in_tok": 586, "out_tok": 234} {"draft_name": "draft-zlgsgl-rtgwg-agents-networking-framework", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft proposes a networking framework for AI agents in enterprise and broadband environments, defining core components and their interactions. The framework aims to enable agent-to-agent communication and coordination in network management scenarios.\",\"n\":2,\"nn\":\"Incremental work extending existing network management concepts to agent-based systems without significant architectural innovation\",\"m\":2,\"mn\":\"Early stage draft with basic framework outline but lacks detailed protocol specifications, implementation details, or concrete examples\",\"o\":3,\"on\":\"Shares conceptual overlap with existing network management frameworks and some agent communication proposals, but focuses on specific enterprise/broadband domain\",\"mo\":2,\"mon\":\"Single draft submission with no evidence of working group adoption or broader community engagement yet\",\"r\":4,\"rn\":\"Directly addresses AI agent networking which is highly relevant to current industry needs for autonomous network operations\",\"c\":[\"A2A protocols\",\"Autonomous netops\",\"Agent discovery/reg\"]}", "in_tok": 487, "out_tok": 199} {"draft_name": "draft-williams-netmod-lm-hierarchy-topology", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft defines a YANG data model for hierarchical topology coordination between language models of different sizes (tiny, small, large) in distributed inference scenarios. It proposes a standards-based approach for orchestrating secure LM interactions across heterogeneous edge-to-cloud environments with request escalation and decentralized validation.\",\n \"n\": 4,\n \"nn\": \"Notable originality in applying network topology modeling to distributed LM coordination with formal YANG specification\",\n \"m\": 3,\n \"mn\": \"Defined protocol mechanism with hierarchical architecture but likely needs more implementation details\",\n \"o\": 2,\n \"on\": \"Minor similarities to distributed AI serving approaches but unique focus on hierarchical LM topology modeling\",\n \"mo\": 2,\n \"mon\": \"Single revision from 2026 with no visible community engagement yet\",\n \"r\": 5,\n \"rn\": \"Core AI agent topic addressing distributed inference coordination between multiple language models\",\n \"c\": [\"A2A protocols\", \"Model serving/inference\", \"Agent identity/auth\", \"Data formats/interop\"]\n}", "in_tok": 604, "out_tok": 258} {"draft_name": "draft-tiloca-lake-app-profiles", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This document defines mechanisms for coordinating and discovering EDHOC application profiles, including a CBOR-based canonical representation for describing and distributing these profiles. It addresses the need for standardized ways to agree on EDHOC parameters that are currently negotiated out-of-band.\",\"n\":2,\"nn\":\"Incremental work that systematizes existing ad-hoc practices for EDHOC profile coordination\",\"m\":4,\"mn\":\"Detailed specification with CBOR schemas, discovery mechanisms, and concrete well-known profiles\",\"o\":2,\"on\":\"Minor overlap with general service discovery work but addresses EDHOC-specific coordination needs\",\"mo\":3,\"mon\":\"Active development with regular revisions but limited to specialized EDHOC community\",\"r\":1,\"rn\":\"Pure cryptographic protocol work for key exchange coordination, not related to AI or agents\",\"c\":[\"Other AI/agent\"]}", "in_tok": 634, "out_tok": 198} {"draft_name": "draft-mzsg-rtgwg-agent-cross-device-comm-framework", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"Proposes a framework for AI agents deployed on network devices to communicate and collaborate across devices for tasks like network measurement and troubleshooting. Analyzes requirements for inter-agent communication protocols in networking contexts.\",\n \"n\": 3,\n \"nn\": \"Useful contribution addressing the specific need for cross-device AI agent coordination in networking, though builds on established multi-agent communication concepts\",\n \"m\": 2,\n \"mn\": \"Early stage work that appears to focus on framework definition and requirements analysis rather than detailed protocol specification\",\n \"o\": 2,\n \"on\": \"Minor overlap with general multi-agent systems work, but focuses specifically on network device deployment which is relatively unique\",\n \"mo\": 2,\n \"mon\": \"Single draft submission with no clear indication of working group adoption or broader community engagement\",\n \"r\": 5,\n \"rn\": \"Directly addresses core AI agent networking challenges including cross-device communication, collaboration protocols, and agent coordination\",\n \"c\": [\"A2A protocols\", \"Autonomous netops\", \"Agent discovery/reg\"]\n}", "in_tok": 559, "out_tok": 254} {"draft_name": "draft-coetzee-oauth-spt-txn-tokens", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"Specifies the Sovereign Policy Token Transaction (SPT-Txn) Framework that extends IETF Transaction Tokens with Capability Acquisition Tokens for fine-grained, auditable authorization across trust boundaries. Designed for regulated financial infrastructure, AI agent systems, and cross-organizational transaction chains requiring cryptographic guarantees of human-origin authorization.\",\n \"n\": 4,\n \"nn\": \"Notable originality in combining capability-based authorization with transaction tokens, cryptographic binding of human identity commitments across delegation chains, and offline verification without issuer interaction\",\n \"m\": 4,\n \"mn\": \"Detailed specification with concrete framework extending RFC 9700, includes specific token formats (CATs), cryptographic mechanisms, and audit trail capabilities\",\n \"o\": 2,\n \"on\": \"Minor similarities to existing OAuth and transaction token work, but unique approach to capability binding and human-origin identity propagation in multi-hop scenarios\",\n \"mo\": 2,\n \"mn\": \"Single revision from future date (2026-04-27) suggests early stage, no evidence of working group adoption or broader community engagement yet\",\n \"r\": 5,\n \"rn\": \"Core AI agent topic addressing critical authorization and policy enforcement needs for agentic systems operating across organizational boundaries with human oversight requirements\",\n \"c\": [\"Agent identity/auth\", \"Policy/governance\", \"A2A protocols\"]\n}", "in_tok": 695, "out_tok": 325} {"draft_name": "draft-gaikwad-aps-profile", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft defines an experimental Agent Persistent State (APS) Profile to address storage requirements for autonomous agents that maintain durable state like user preferences, embeddings, and audit traces. It proposes a vendor-neutral storage service class with emphasis on compliance, cryptographic erasure capabilities, and integration with container orchestration systems.\",\n \"n\": 4,\n \"nn\": \"Notable originality in recognizing and addressing the 'layer mismatch' between agent storage needs and current infrastructure, with novel focus on compliance-oriented storage including cryptographic erasure\",\n \"m\": 3,\n \"mn\": \"Well-defined protocol with clear usage classes and schema, though lacks detailed implementation examples and test vectors for full implementation readiness\",\n \"o\": 2,\n \"on\": \"Minor similarities to general storage standards but unique in its agent-specific focus and compliance requirements\",\n \"mo\": 2,\n \"mn\": \"Single revision draft with experimental status, no indication of working group adoption or broader community engagement yet\",\n \"r\": 5,\n \"rn\": \"Core AI agent infrastructure topic addressing fundamental storage requirements for autonomous agents with persistent state\",\n \"c\": [\"Data formats/interop\", \"Agent identity/auth\", \"Policy/governance\"]\n}", "in_tok": 736, "out_tok": 288} {"draft_name": "draft-bates-atp", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"ATP defines a cryptographically verifiable DAG protocol for tracking agent transactions with tamper-evident signed causality chains. The protocol aims to provide accountability and auditable lineage across distributed agentic systems through lightweight, transport-independent mechanisms.\",\n \"n\": 4,\n \"nn\": \"Notable application of DAG cryptographic verification specifically to agent transaction tracking - bridges blockchain concepts with AI agent accountability in a novel way\",\n \"m\": 4,\n \"mn\": \"65 pages suggests detailed specification with comprehensive protocol definition, though implementation readiness unclear without test vectors mentioned\",\n \"o\": 2,\n \"on\": \"Some conceptual overlap with general cryptographic audit trails and blockchain structures, but specific application to agent transactions appears unique\",\n \"mo\": 2,\n \"mon\": \"Future date (2026) and lack of WG context suggests early individual submission without established community momentum\",\n \"r\": 5,\n \"rn\": \"Directly addresses core challenges in agent systems - accountability, provenance, and verifiable history are fundamental concerns for AI agent deployment\",\n \"c\": [\"A2A protocols\", \"Agent identity/auth\", \"AI safety/alignment\"]\n}", "in_tok": 555, "out_tok": 273} {"draft_name": "draft-dunbar-onions-ac-te-applicability", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This document explores applying existing IETF YANG data models for Attachment Circuit and Traffic Engineering to support dynamic AI model placement at edge cloud sites. It maps a use case of optimal edge location selection for real-time AI inference from street cameras across network segments to evaluate model applicability.\",\n \"n\": 2,\n \"nn\": \"Incremental application of existing YANG models to edge AI placement - useful but not particularly novel\",\n \"m\": 3,\n \"mn\": \"Defines specific protocol application with concrete use case mapping, though lacks implementation details\",\n \"o\": 3,\n \"on\": \"Shares concepts with existing edge computing and network management work, moderate overlap expected\",\n \"mo\": 2,\n \"mon\": \"Single revision from October 2025, too recent to assess broader momentum\",\n \"r\": 4,\n \"rn\": \"Directly relevant to AI agent infrastructure, focusing on network-aware AI model placement and service optimization\",\n \"c\": [\"Model serving/inference\", \"Autonomous netops\", \"ML traffic mgmt\"]\n}", "in_tok": 629, "out_tok": 252} {"draft_name": "draft-ahn-nmrg-5g-security-i2nsf-framework", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft proposes an integrated security framework for 5G networks that uses Intent-Based Networking to translate high-level security intents into enforceable policies via I2NSF architecture. The system distributes policy enforcement across network and application levels with adaptive response to dynamic edge conditions.\",\n \"n\": 2,\n \"nn\": \"Incremental combination of existing technologies (I2NSF, IBN, 5G NEF) without novel algorithmic or architectural contributions\",\n \"m\": 2,\n \"mn\": \"Early sketch level - describes framework components and interactions but lacks detailed protocol specifications or implementation details\",\n \"o\": 4,\n \"on\": \"Significant overlap with existing I2NSF work and intent-based security management approaches in SDN/NFV contexts\",\n \"mo\": 2,\n \"mon\": \"Single revision from NMRG with limited evidence of broader community engagement or implementation activity\",\n \"r\": 3,\n \"rn\": \"Partially relevant through intent-based networking aspects, but primarily focused on 5G security rather than AI agents\",\n \"c\": [\"Autonomous netops\", \"Policy/governance\"]\n}", "in_tok": 670, "out_tok": 272} {"draft_name": "draft-zhao-ccamp-actn-optical-network-agent", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft proposes integrating AI-based Network Management Agents (NMA) into ACTN optical networks to achieve higher autonomy levels. It extends the ACTN architecture with agent-based components for autonomous perception, decision-making, and execution in optical network operations.\",\n \"n\": 3,\n \"nn\": \"Useful integration of AI agents into established ACTN framework, though applying AI to network management is well-explored territory\",\n \"m\": 2,\n \"mn\": \"Early architectural sketch with high-level components and use cases, lacks detailed protocols and implementation specifics\",\n \"o\": 3,\n \"on\": \"Shares autonomous network concepts with other drafts but focuses specifically on optical networks and ACTN integration\",\n \"mo\": 2,\n \"mon\": \"Single revision from February 2026, no clear indication of working group adoption or community engagement\",\n \"r\": 4,\n \"rn\": \"Directly addresses AI agent integration in network management with specific focus on autonomous operations and agent-based decision making\",\n \"c\": [\"Autonomous netops\", \"A2A protocols\"]\n}", "in_tok": 724, "out_tok": 264} {"draft_name": "draft-jurkovikj-httpapi-agentic-state", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"Defines Agentic State Transfer (AST), an HTTP profile for maintaining consistency across multiple representations of the same resource using semantic validators and optimistic concurrency control. Addresses the problem where ETags are representation-specific rather than reflecting the underlying logical state of a resource.\",\n \"n\": 4,\n \"nn\": \"Notable originality in addressing representation-level vs. logical-state consistency in HTTP, with the concept of Canonical Resource State being a fresh approach to multi-representation synchronization\",\n \"m\": 3,\n \"mn\": \"Defines clear protocol mechanisms with semantic validators and mandatory optimistic concurrency control, though could benefit from more implementation examples and edge case handling\",\n \"o\": 2,\n \"on\": \"Minor similarities to existing HTTP extension drafts but the specific focus on cross-representation state consistency is relatively unique\",\n \"mo\": 2,\n \"mn\": \"Single revision draft with no clear working group adoption or community discussion visible\",\n \"r\": 4,\n \"rn\": \"Directly relevant to AI agents which often need to interact with resources through multiple representations and require consistent state management for reliable operation\",\n \"c\": [\"A2A protocols\", \"Data formats/interop\"]\n}", "in_tok": 697, "out_tok": 283} {"draft_name": "draft-sharif-agent-identity-framework", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft proposes a five-layer framework (identity, authorization, attestation, evidence, trust) to address gaps in Internet protocols for autonomous AI agent authentication and authorization. It serves as an informational guide for future standards work rather than defining specific wire protocols.\",\n \"n\": 4,\n \"nn\": \"Notable originality in systematically decomposing agent trust into distinct layers and identifying specific gaps between human-centric Internet protocols and autonomous agent requirements\",\n \"m\": 3,\n \"mn\": \"Well-defined framework with clear layer separation and gap analysis, but remains at conceptual level without detailed protocol specifications or implementation guidance\",\n \"o\": 2,\n \"on\": \"Minor similarities to general identity/authentication work, but the autonomous agent focus and five-layer decomposition appear relatively unique\",\n \"mo\": 2,\n \"mon\": \"Single revision with future date suggests early stage work without established community engagement or working group adoption\",\n \"r\": 5,\n \"rn\": \"Core AI agent topic directly addressing fundamental trust and identity challenges for autonomous systems operating on the Internet\",\n \"c\": [\"Agent identity/auth\", \"Policy/governance\"]\n}", "in_tok": 708, "out_tok": 269} {"draft_name": "draft-cosmos-protocol-specification", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"The Cosmos Protocol defines a comprehensive badge-based identity and communication system with 8 foundational components including cryptographic identity, trust scoring, post-quantum encryption, and semantic messaging. It aims to enable trust-native communication without requiring blockchain infrastructure while supporting passwordless authentication and decentralized identity management.\",\n \"n\": 4,\n \"nn\": \"Notable integration of badge-based identity with trust scoring, post-quantum encryption, and semantic messaging in a unified protocol avoiding blockchain dependencies\",\n \"m\": 2,\n \"mn\": \"Abstract cuts off mid-sentence and appears to be early draft with undefined RFC references and incomplete component descriptions\",\n \"o\": 3,\n \"on\": \"Shares identity and authentication concepts with existing decentralized identity work, but badge-based approach and trust-native design represents distinct implementation\",\n \"mo\": 2,\n \"mon\": \"Single draft submission with no apparent working group adoption or community discussion visible\",\n \"r\": 4,\n \"rn\": \"Directly relevant to AI agents through identity management, trust scoring, semantic messaging grammar, and agent compatibility features\",\n \"c\": [\"Agent identity/auth\", \"A2A protocols\", \"Data formats/interop\"]\n}", "in_tok": 878, "out_tok": 281} {"draft_name": "draft-sharif-agent-payment-trust", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"Defines a trust scoring and identity verification protocol for autonomous AI agents conducting financial transactions, featuring five-dimension trust metrics, cryptographic identities, and spend limits. Complements existing secure messaging protocols to enable trustworthy AI agent payment ecosystems.\",\n \"n\": 4,\n \"nn\": \"Notable originality in combining multi-dimensional trust scoring with behavioral anomaly detection specifically for AI agent financial transactions - addresses emerging needs not well covered by existing identity/payment standards\",\n \"m\": 4,\n \"mn\": \"Detailed specification with concrete trust scoring dimensions, cryptographic identity mechanisms using ECDSA P-256, spend limit tiers, and public API definitions - appears implementation-ready\",\n \"o\": 2,\n \"on\": \"Minor similarities to general identity and payment protocols, but specific focus on AI agent trust scoring and behavioral analysis represents a distinct approach\",\n \"mo\": 3,\n \"mon\": \"Active development as part of broader AI agent protocol suite, but unclear level of community adoption or working group interest at this early stage\",\n \"r\": 5,\n \"rn\": \"Core AI agent topic directly addressing critical trust and financial security needs for autonomous agent ecosystems - highly relevant to emerging AI agent economy\",\n \"c\": [\"Agent identity/auth\", \"A2A protocols\", \"AI safety/alignment\"]\n}", "in_tok": 685, "out_tok": 306} {"draft_name": "draft-ietf-vwrap-authentication", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft defines authentication mechanisms for the Virtual World Region Agent Protocol (VWRAP), establishing trust models for client applications to verify user credentials before granting control over virtual world agents and digital assets. It focuses on identity verification and capability granting in virtual world environments.\",\n \"n\": 2,\n \"nn\": \"Incremental work applying standard authentication patterns to virtual world agents - not particularly novel but addresses a specific domain need\",\n \"m\": 4,\n \"mn\": \"Detailed specification with clear protocol definitions, trust model descriptions, and sufficient technical depth for implementation\",\n \"o\": 3,\n \"on\": \"Shares authentication and identity concepts with existing IETF identity work, but specific to virtual world agent domain\",\n \"mo\": 2,\n \"mon\": \"VWRAP working group has limited recent activity, and virtual world standardization efforts have historically struggled with adoption\",\n \"r\": 4,\n \"rn\": \"Directly relevant to agent identity and authentication, addressing how users authenticate to control virtual agents and their digital assets\",\n \"c\": [\"Agent identity/auth\", \"A2A protocols\"]\n}", "in_tok": 518, "out_tok": 264} {"draft_name": "draft-yuan-rtgwg-security-agent-usecase", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft proposes AI Network Security Agents for routers that use machine learning and behavioral analysis to provide adaptive, automated security responses. It outlines four use cases: dynamic defense against unknown threats, ACL optimization, configuration security, and collaborative defense across network devices.\",\n \"n\": 2,\n \"nn\": \"Incremental application of well-established AI/ML security concepts to network routing devices, without introducing novel technical approaches\",\n \"m\": 2,\n \"mn\": \"Early sketch that describes use cases and high-level capabilities but lacks detailed protocols, mechanisms, or implementation specifics\",\n \"o\": 4,\n \"on\": \"Significant overlap with existing AI-driven network security, autonomous network operations, and ML-based threat detection work in IETF and broader industry\",\n \"mo\": 1,\n \"mn\": \"Future-dated draft (2026) suggests this is speculative work with no current community activity or momentum\",\n \"r\": 4,\n \"rn\": \"Directly relevant as it focuses on AI agents for network security with autonomous decision-making capabilities\",\n \"c\": [\"Autonomous netops\", \"ML traffic mgmt\", \"AI safety/alignment\"]\n}", "in_tok": 672, "out_tok": 279} {"draft_name": "draft-mao-rtgwg-apn-framework-for-ioa", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft proposes applying the Application-aware Networking (APN) framework to Internet of Agents (IoA) scenarios, where AI agents from different platforms need to communicate with varying network requirements. It analyzes the necessity of providing fine-grained network services for AI agent interactions.\",\n \"n\": 2,\n \"nn\": \"Incremental application of existing APN framework to AI agent scenarios - straightforward extension rather than novel approach\",\n \"m\": 2,\n \"mn\": \"Early sketch at 11 pages analyzing necessity but likely lacking detailed protocol mechanisms or implementation specifics\",\n \"o\": 3,\n \"on\": \"Shares APN concepts with existing networking drafts and IoA concepts with other AI agent networking proposals\",\n \"mo\": 2,\n \"mon\": \"Future date (2026-05-05) suggests single early revision without established community traction yet\",\n \"r\": 4,\n \"rn\": \"Directly addresses AI agent networking requirements and inter-agent communication challenges in IoA scenarios\",\n \"c\": [\"A2A protocols\", \"ML traffic mgmt\"]\n}", "in_tok": 618, "out_tok": 261} {"draft_name": "draft-howe-vcon-agent-session", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft defines an agent_session extension for vCon (Virtualized Conversations) to capture autonomous AI agents' internal sessions including prompts, tool calls, reasoning traces, and file artifacts alongside human conversations. It leverages existing vCon infrastructure for identity management, lifecycle handling, and cryptographic signing without requiring new top-level fields.\",\"n\":3,\"nn\":\"Useful contribution that adapts existing conversation recording format for AI agent sessions, though the core concept of structured agent trace capture is somewhat expected\",\"m\":4,\"mn\":\"Detailed specification with clear data model, CDDL schema references, and concrete examples of how agent sessions map to vCon structures\",\"o\":2,\"on\":\"Minor similarities to conversation recording formats but unique in its specific mapping of agent internals to vCon model\",\"mo\":3,\"mon\":\"Active development with recent date and part of broader vCon ecosystem, though unclear if adopted by working group\",\"r\":5,\"rn\":\"Directly addresses core need for standardized AI agent session recording, tracing, and auditability in conversational AI systems\",\"c\":[\"Data formats/interop\",\"Human-agent interaction\",\"Policy/governance\"]}", "in_tok": 807, "out_tok": 255} {"draft_name": "draft-bernardos-cats-isac-uc", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft proposes Integrated Sensing and Communications (ISAC) as a use case for Computing-Aware Traffic Steering (CATS), where wireless networks jointly optimize sensing and communication functions. The document aims to identify challenges and requirements for ISAC scenarios in autonomous systems, smart cities, and industrial automation.\",\n \"n\": 2,\n \"nn\": \"Incremental application of existing CATS framework to ISAC domain without novel technical contributions\",\n \"m\": 2,\n \"mn\": \"Early sketch that presents use case scenarios but lacks detailed protocol mechanisms or specifications\",\n \"o\": 3,\n \"on\": \"Shares conceptual overlap with other CATS use case documents and ISAC research literature\",\n \"mo\": 3,\n \"mon\": \"Part of active CATS working group development with recent publication date suggesting ongoing work\",\n \"r\": 3,\n \"rn\": \"Partially relevant as it addresses autonomous systems and intelligent traffic steering but focuses more on wireless sensing than core AI agent protocols\",\n \"c\": [\"ML traffic mgmt\", \"Autonomous netops\", \"Other AI/agent\"]\n}", "in_tok": 601, "out_tok": 264} {"draft_name": "draft-huang-rats-agentic-eat-cap-attest", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft extends the Entity Attestation Token (EAT) RFC to support capability attestation for agentic AI systems, introducing new claims for asserting agent functional, reasoning, and operational capabilities. It aims to enable trustworthy verification of autonomous agent capabilities in decentralized environments.\",\n \"n\": 4,\n \"nn\": \"Notable application of existing attestation framework to emerging agentic AI domain with specific capability-focused extensions\",\n \"m\": 2,\n \"mn\": \"At 6 pages, likely provides early conceptual framework but insufficient detail for implementation-ready specification\",\n \"o\": 2,\n \"on\": \"Builds on established EAT RFC but applies it to relatively new agentic AI attestation domain with minimal existing overlap\",\n \"mo\": 2,\n \"mn\": \"Recent draft from December 2025 but no indication of working group adoption or broader community engagement yet\",\n \"r\": 5,\n \"rn\": \"Directly addresses core challenge of verifying and attesting capabilities of autonomous AI agents in decentralized systems\",\n \"c\": [\"Agent identity/auth\", \"AI safety/alignment\", \"Data formats/interop\"]\n}", "in_tok": 588, "out_tok": 278} {"draft_name": "draft-huang-acme-scalable-agent-enrollment", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft proposes extending certificate enrollment protocols (SCEP/EST) for AI agent identity using two scalable models: Zero-Knowledge Proofs for private continuous attestation, and one-time host bootstrapping for trusted endorsement. The approach addresses performance bottlenecks of per-enrollment verification while maintaining high-assurance identity for agentic AI systems.\",\n \"n\": 4,\n \"nn\": \"Notable approach combining ZKP with traditional PKI for AI agent identity, and the dual-model framework addressing both privacy and scalability is original\",\n \"m\": 2,\n \"mn\": \"Early sketch level - presents models and concepts but lacks detailed protocol specifications, message formats, or concrete implementation guidance\",\n \"o\": 2,\n \"on\": \"Some conceptual overlap with general attestation and PKI work, but the specific AI agent focus and dual-model approach are fairly unique\",\n \"mo\": 2,\n \"mn\": \"Single revision from December 2024, too early to assess broader community interest or working group adoption\",\n \"r\": 5,\n \"rn\": \"Core AI agent topic directly addressing fundamental identity and authentication needs for autonomous AI systems at scale\",\n \"c\": [\"Agent identity/auth\", \"A2A protocols\"]\n}", "in_tok": 655, "out_tok": 294} {"draft_name": "draft-zhao-nmrg-ai-agent-for-dtn", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"Proposes integrating AI agents into Digital Twin Network (DTN) architecture to enable autonomous network management capabilities. Extends traditional DTN by adding AI agents at each component level for more intelligent and adaptive operations.\",\"n\":2,\"nn\":\"Incremental combination of existing AI agent concepts with digital twin networking - straightforward extension rather than novel approach\",\"m\":2,\"mn\":\"Early architectural sketch without detailed protocols, mechanisms, or implementation specifics\",\"o\":4,\"on\":\"Significant overlap with existing digital twin networking and AI-driven network management literature\",\"mo\":2,\"mon\":\"Single revision with limited evidence of community engagement or working group interest\",\"r\":4,\"rn\":\"Directly relevant to AI agents in networking context, focusing on autonomous network operations\",\"c\":[\"Autonomous netops\",\"Other AI/agent\"]}", "in_tok": 512, "out_tok": 176} {"draft_name": "draft-li-cdn-node-selection-enhancement", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft proposes enhancing CDN node selection by combining anycast routing with QUIC transport to enable dynamic route advertisement based on real-time load and application requirements. The solution allows clients to signal service needs and enables server-initiated redirection using QUIC connection migration for more intelligent CDN performance.\",\"n\":2,\"nn\":\"Incremental combination of existing technologies (anycast, BGP, QUIC) for CDN optimization - not particularly novel but useful integration\",\"m\":2,\"mn\":\"Early sketch level - abstract describes concepts but lacks detailed protocol mechanisms, message formats, or implementation specifics\",\"o\":3,\"on\":\"Shares concepts with existing CDN optimization and anycast routing work, plus QUIC connection migration proposals, but specific combination appears distinct\",\"mo\":1,\"mon\":\"Single draft submission with no apparent working group activity or broader community engagement visible\",\"r\":1,\"rn\":\"This is purely about CDN infrastructure optimization and network protocols - no AI or agent-related content despite the rating request\"}", "in_tok": 702, "out_tok": 219} {"draft_name": "draft-song-lake-ra", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft specifies how to integrate remote attestation capabilities into the EDHOC lightweight key exchange protocol using the RATS architecture. It enables devices to verify each other's trustworthiness during the authentication process, combining cryptographic key establishment with hardware-based attestation.\",\"n\":3,\"nn\":\"Useful integration of existing attestation and key exchange protocols, addresses practical need for trusted communication establishment\",\"m\":4,\"mn\":\"Detailed specification with concrete protocol flows, message formats, and security considerations, appears implementation-ready\",\"o\":2,\"on\":\"Builds on established EDHOC and RATS foundations but presents distinct integration approach\",\"mo\":2,\"mon\":\"Recent draft with limited revision history, unclear community adoption status\",\"r\":2,\"rn\":\"Primarily focused on device attestation and secure communications rather than AI agents specifically, though could support agent authentication\",\"c\":[\"Agent identity/auth\"]}", "in_tok": 510, "out_tok": 195} {"draft_name": "draft-hu-neotec-usecases-notc", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft presents two telco cloud use cases involving AI-based video recognition for smart cities and dynamic high-bandwidth transport. It focuses on orchestrating heterogeneous computing resources (GPU/NPU) with network controllers for optimized performance and QoS.\",\n \"n\": 2,\n \"nn\": \"Incremental application of existing AI orchestration concepts to telco cloud environments, combining well-known elements like distributed inference and elastic provisioning\",\n \"m\": 2,\n \"mn\": \"Early stage use case descriptions without concrete protocols, mechanisms, or implementation details - primarily problem statement and requirements\",\n \"o\": 4,\n \"on\": \"Significant overlap with existing network function virtualization, cloud orchestration, and AI inference deployment literature and standards\",\n \"mo\": 2,\n \"mon\": \"Single draft submission with no evidence of working group adoption or broader community engagement in this specific formulation\",\n \"r\": 3,\n \"rn\": \"Partially relevant as it addresses AI deployment in network infrastructure, but focuses more on traditional orchestration than AI agent behaviors or protocols\",\n \"c\": [\"Autonomous netops\", \"Model serving/inference\"]\n}", "in_tok": 597, "out_tok": 271} {"draft_name": "draft-cui-ai-agent-task", "model": "claude-sonnet-4-20250514", "raw": "```json\n{\n \"s\": \"This draft outlines requirements for AI agent protocols to coordinate tasks autonomously across large-scale networks. It focuses on dynamic task discovery, negotiation, and context-aware scheduling capabilities.\",\n \"n\": 3,\n \"nn\": \"Addresses important coordination challenges but builds on established multi-agent systems concepts\",\n \"m\": 2,\n \"mn\": \"Early requirements document without detailed protocol specifications or concrete mechanisms\",\n \"o\": 3,\n \"on\": \"Shares coordination concepts with existing multi-agent protocol work but focuses specifically on network-scale deployment\",\n \"mo\": 2,\n \"mon\": \"Single revision with no evidence of working group interest or community engagement yet\",\n \"r\": 5,\n \"rn\": \"Directly addresses core AI agent coordination challenges essential for distributed agent systems\",\n \"c\": [\"A2A protocols\", \"Agent discovery/reg\", \"Autonomous netops\"]\n}\n```", "in_tok": 520, "out_tok": 222} {"draft_name": "draft-nelson-agent-delegation-receipts", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft defines a cryptographic protocol where users sign authorization receipts before AI agents execute actions, removing operators as trusted parties by using append-only logs and user private keys as sole signing authorities. The protocol aims to provide transparent, verifiable delegation of authority to AI agents with defined scope and time boundaries.\",\"n\":4,\"nn\":\"Notable originality in combining cryptographic receipts with AI agent authorization, creating a novel trust model that removes operator intermediaries\",\"m\":4,\"mn\":\"Detailed 45-page specification with defined protocol elements, cryptographic primitives, and specific implementation guidance\",\"o\":2,\"on\":\"Minor conceptual overlap with general authorization frameworks but unique application to AI agent delegation with cryptographic receipts\",\"mo\":3,\"mon\":\"Active development indicated by 2026 date, though specific working group status unclear\",\"r\":5,\"rn\":\"Core AI agent topic addressing fundamental authorization and trust challenges in agent deployments\",\"c\":[\"Agent identity/auth\",\"AI safety/alignment\",\"Policy/governance\"]}", "in_tok": 560, "out_tok": 217} {"draft_name": "draft-song-oauth-ai-agent-collaborate-authz", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft extends OAuth 2.0 to enable authorization for multi-AI agent collaboration scenarios where a lead agent coordinates sub-agents. It adds fields to tokens and message flows to allow sub-agents to execute tasks as a group while maintaining OAuth 2.0 compatibility.\",\n \"n\": 3,\n \"nn\": \"Useful contribution applying established OAuth patterns to emerging multi-agent AI scenarios, though the core concept of delegated authorization is well-established\",\n \"m\": 2,\n \"mn\": \"Early sketch level - lacks detailed protocol specifications, message formats, security considerations, and concrete implementation guidance\",\n \"o\": 3,\n \"on\": \"Shares authorization delegation concepts with existing OAuth extensions and multi-party authentication schemes, but specific AI agent application is less covered\",\n \"mo\": 2,\n \"mon\": \"Single revision with future date suggests early stage work without demonstrated community engagement or working group adoption\",\n \"r\": 4,\n \"rn\": \"Directly addresses a relevant emerging challenge in AI agent ecosystems - how to handle authorization in collaborative multi-agent workflows\",\n \"c\": [\"Agent identity/auth\", \"A2A protocols\"]\n}", "in_tok": 577, "out_tok": 272} {"draft_name": "draft-yu-ai-agent-use-cases-in-6g", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft outlines use cases and requirements for AI agents in 6G networks, building on 3GPP technical reports. It focuses on operator perspectives for integrating AI agents into next-generation mobile network infrastructure.\",\n \"n\": 2,\n \"nn\": \"Incremental work that primarily references existing 3GPP studies without introducing fundamentally new concepts for AI agents in telecommunications\",\n \"m\": 2,\n \"mn\": \"Early sketch that identifies use cases and high-level requirements but lacks detailed protocols, mechanisms, or concrete technical specifications\",\n \"o\": 4,\n \"on\": \"Significant overlap with existing 3GPP TR 22.870 and other telecommunications AI integration work, with limited unique IETF-specific contributions\",\n \"mo\": 2,\n \"mon\": \"Single revision draft with no clear indication of working group interest or broader community engagement in IETF context\",\n \"r\": 4,\n \"rn\": \"Directly relevant to AI agents topic, specifically addressing their integration into network infrastructure and telecommunications systems\",\n \"c\": [\"Autonomous netops\", \"ML traffic mgmt\", \"Policy/governance\"]\n}", "in_tok": 535, "out_tok": 271} {"draft_name": "draft-bernardos-anima-fog-monitoring", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft proposes using GRASP (Generic Autonomic Signaling Protocol) to autonomically setup monitoring agents in fog computing environments. It addresses the challenge of dynamically deploying monitoring infrastructure across volatile and mobile fog resources to enable predictive orchestration.\",\"n\":2,\"nn\":\"Incremental application of existing GRASP protocol to fog monitoring - straightforward extension of autonomic networking concepts\",\"m\":2,\"mn\":\"Early specification with basic GRASP usage outlined but lacks detailed protocol flows, error handling, and implementation specifics\",\"o\":3,\"on\":\"Shares autonomic networking concepts with other ANIMA drafts and overlaps with general fog computing monitoring approaches\",\"mo\":2,\"mon\":\"Single revision in ANIMA area with limited evidence of broader community engagement or working group adoption\",\"r\":4,\"rn\":\"Directly relevant as it focuses on autonomous agent deployment and management for network monitoring in dynamic environments\",\"c\":[\"Autonomous netops\",\"Agent discovery/reg\"]}", "in_tok": 766, "out_tok": 212} {"draft_name": "draft-rosenberg-cheq", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"CHEQ proposes a protocol for requiring human confirmation of AI agent actions before execution, using signed objects that can be integrated into existing agent protocols like MCP and A2A. The protocol aims to prevent AI hallucinations from causing unwanted actions while protecting human privacy during the confirmation process.\",\n \"n\": 3,\n \"nn\": \"Useful contribution addressing the important problem of AI agent oversight, though human-in-the-loop concepts are well-established in other domains\",\n \"m\": 3,\n \"mn\": \"Defines core protocol mechanisms and integration approach with existing protocols, but likely needs more detailed specification and examples\",\n \"o\": 2,\n \"on\": \"While addressing AI agent control, the specific approach of confirmation protocols with signed objects appears to have minimal overlap with existing IETF work\",\n \"mo\": 2,\n \"mon\": \"Single revision from July 2025 with no clear indication of working group adoption or broader community engagement yet\",\n \"r\": 5,\n \"rn\": \"Directly addresses core AI agent safety and control mechanisms, highly relevant to AI agent standardization efforts\",\n \"c\": [\"Human-agent interaction\", \"AI safety/alignment\", \"A2A protocols\"]\n}", "in_tok": 632, "out_tok": 286} {"draft_name": "draft-mallick-muacp", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft specifies \u00b5ACP, a resource-efficient messaging protocol designed specifically for autonomous agents running on constrained IoT and Edge devices. It provides deterministic resource bounds with an 8-byte header, TLV extensibility, and mandatory OSCORE security for agent coordination in resource-limited environments.\",\"n\":4,\"nn\":\"Notable originality in designing agent communication specifically for resource-constrained devices with deterministic bounds\",\"m\":4,\"mn\":\"Detailed specification with concrete resource limits, header format, and security requirements\",\"o\":2,\"on\":\"Some conceptual overlap with constrained protocols but unique focus on agent communication\",\"mo\":2,\"mon\":\"Single revision, appears to be individual submission without clear WG adoption\",\"r\":5,\"rn\":\"Core AI agent topic addressing fundamental communication needs for autonomous agents\",\"c\":[\"A2A protocols\",\"Agent identity/auth\"]}", "in_tok": 634, "out_tok": 191} {"draft_name": "draft-li-dmsc-mcps-agw", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft specifies a protocol suite using Agent Gateways as control-plane entities for managing distributed agent collaboration, including registration, authentication, and semantic routing while preserving direct peer-to-peer interactions. The approach aims to enable scalable and secure multi-agent coordination across heterogeneous networks.\",\n \"n\": 3,\n \"nn\": \"Useful contribution combining gateway-based control plane with direct agent interaction, though builds on established distributed systems patterns\",\n \"m\": 2,\n \"mn\": \"Early specification stage with basic architectural concepts but likely lacking detailed protocol mechanics and implementation specifics\",\n \"o\": 3,\n \"on\": \"Shares concepts with existing agent discovery and multi-agent coordination work but has distinct gateway-centric approach\",\n \"mo\": 2,\n \"mon\": \"Single draft with recent date suggests early development phase, no clear working group adoption signals\",\n \"r\": 5,\n \"rn\": \"Directly addresses core AI agent infrastructure needs for distributed collaboration, authentication, and semantic routing\",\n \"c\": [\"A2A protocols\", \"Agent identity/auth\", \"Agent discovery/reg\"]\n}", "in_tok": 546, "out_tok": 265} {"draft_name": "draft-cowles-aee", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"AEE defines a minimal JSON envelope format with 14 fields for inter-agent communication, separating message routing concerns from payload semantics. The specification provides message identity, correlation, tracing, and priority handling without prescribing transport mechanisms or domain-specific meanings.\",\n \"n\": 3,\n \"nn\": \"Useful contribution that addresses envelope standardization for agent communication, though building on established messaging patterns\",\n \"m\": 4,\n \"mn\": \"Detailed specification with well-defined fields, validity rules, conformance levels, and clear examples ready for implementation\",\n \"o\": 3,\n \"on\": \"Shares messaging envelope concepts with general distributed systems work but focuses specifically on agent communication needs\",\n \"mo\": 2,\n \"mon\": \"Single revision with future date suggests early stage without demonstrated community engagement or working group adoption\",\n \"r\": 5,\n \"rn\": \"Core AI agent topic addressing fundamental interoperability challenge for autonomous agent communication and workflow coordination\",\n \"c\": [\"A2A protocols\", \"Data formats/interop\"]\n}", "in_tok": 635, "out_tok": 252} {"draft_name": "draft-yang-dmsc-ioa-task-protocol", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft defines the Internet of Agents Task Protocol (IoA Task Protocol) for enabling distributed collaboration among heterogeneous AI agents across diverse network environments. The protocol provides layered architecture with dynamic team formation, adaptive task coordination, and structured messaging to support large-scale intelligent systems in domains like transportation, healthcare, and human-AI teaming.\",\n \"n\": 4,\n \"nn\": \"Notable originality in addressing heterogeneous agent collaboration with focus on dynamic team formation and cross-network deployment, going beyond simple agent-to-agent communication\",\n \"m\": 3,\n \"mn\": \"Defined protocol concept with layered architecture and extensible messaging format, but likely needs more detailed specification and concrete examples for implementation\",\n \"o\": 2,\n \"on\": \"While agent communication protocols exist, the focus on heterogeneous collaboration with dynamic teaming and cross-network deployment shows minor overlap with existing approaches\",\n \"mo\": 2,\n \"mon\": \"Single draft with future date (2026) suggests early stage work without established working group momentum or multiple iterations\",\n \"r\": 5,\n \"rn\": \"Core AI agent topic directly addressing fundamental challenge of multi-agent coordination and collaboration in distributed intelligent systems\",\n \"c\": [\"A2A protocols\", \"Agent discovery/reg\", \"Data formats/interop\", \"Human-agent interaction\"]\n}", "in_tok": 640, "out_tok": 312} {"draft_name": "draft-zw-nmrg-mcp-network-mgmt", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft extends the Model Context Protocol (MCP) to enable network equipment to act as MCP servers for AI-driven network management. It defines new capability tokens, tools, resources, and error codes specifically for network operations while maintaining compatibility with existing MCP implementations.\",\"n\":3,\"nn\":\"Useful adaptation of existing MCP framework to network management domain with domain-specific extensions\",\"m\":4,\"mn\":\"Detailed specification with new capability tokens, tools, resources, prompts and error codes - appears implementation-ready\",\"o\":2,\"on\":\"Builds on MCP framework but applies it to network management domain with minimal overlap to existing work\",\"mo\":2,\"mon\":\"Single revision draft with future date, no indication of working group adoption or community involvement yet\",\"r\":5,\"rn\":\"Directly addresses AI agent interaction with network infrastructure - core AI agent networking topic\",\"c\":[\"A2A protocols\",\"Autonomous netops\",\"Data formats/interop\"]}", "in_tok": 552, "out_tok": 207} {"draft_name": "draft-morrison-substrate-observation", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft proposes substrate-observation as an alternative to envelope coordination for managing concurrent sessions in cross-tool agentic systems. It argues against standardizing certain coordination protocols, instead advocating for substrate-physics primitives based on identity-bound principals and handle namespaces.\",\"n\":4,\"nn\":\"Novel approach to agent coordination that explicitly argues against standardization of envelope formats in favor of substrate-observation primitives\",\"m\":2,\"mn\":\"Early conceptual sketch that relies heavily on companion drafts and lacks detailed protocol specifications or implementation guidance\",\"o\":1,\"on\":\"Unique anti-standardization approach to agent coordination with distinctive substrate-observation concept not seen in other drafts\",\"mo\":2,\"mon\":\"Single revision from 2026 with no evidence of working group adoption or broader community engagement\",\"r\":5,\"rn\":\"Directly addresses core agent coordination challenges in multi-tool agentic systems with identity-bound principals\",\"c\":[\"A2A protocols\",\"Agent identity/auth\"]}", "in_tok": 630, "out_tok": 214} {"draft_name": "draft-liu-rtgwg-agent-gateway-requirements", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft proposes requirements for Agent Gateways to improve scalability, efficiency, and security in agent-to-agent communications. It identifies gaps in current gateway solutions and argues for a new type of gateway entity specifically designed for agent communications.\",\"n\":2,\"nn\":\"Incremental extension of traditional gateway concepts to agent communications, applying familiar networking patterns to AI agents\",\"m\":2,\"mn\":\"Early-stage requirements document without detailed technical specifications or concrete protocol definitions\",\"o\":3,\"on\":\"Shares gateway and agent communication concepts with existing networking and emerging agent infrastructure work\",\"mo\":2,\"mon\":\"Single revision with no clear working group adoption or broader community engagement visible\",\"r\":4,\"rn\":\"Directly addresses agent-to-agent communication infrastructure, which is core to AI agent networking\",\"c\":[\"A2A protocols\",\"Agent discovery/reg\",\"Autonomous netops\"]}", "in_tok": 526, "out_tok": 190} {"draft_name": "draft-zeng-opsawg-applicability-mcp-a2a", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft analyzes NETCONF's limitations for AI-driven network management and proposes MCP and A2A protocols as complementary solutions. It addresses scenarios requiring semantic translation, cross-domain orchestration, and multi-agent consensus that NETCONF cannot handle effectively.\",\"n\":3,\"nn\":\"Useful systematic analysis of protocol gaps with concrete alternative solutions, though building on existing MCP/A2A concepts\",\"m\":2,\"mn\":\"Early stage work with problem analysis and high-level solution mapping but lacks detailed protocol specifications\",\"o\":3,\"on\":\"Shares concepts with network management and agent protocol drafts but focuses on specific NETCONF gap analysis\",\"mo\":2,\"mon\":\"Recent draft with future date suggesting early development, unclear if part of active working group\",\"r\":4,\"rn\":\"Directly addresses AI-driven network management with agent protocols, covering multi-agent consensus and semantic translation\",\"c\":[\"A2A protocols\",\"Autonomous netops\",\"Agent discovery/reg\",\"Data formats/interop\"]}", "in_tok": 587, "out_tok": 220} {"draft_name": "draft-rosenberg-aiproto-cheq", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"CHEQ is a protocol for human-in-the-loop confirmation of AI agent decisions before execution, using signed objects that can be integrated into existing agent protocols like MCP and A2A. It aims to prevent AI hallucinations from causing unwanted actions while protecting human privacy during tool invocation.\",\n \"n\": 3,\n \"nn\": \"Useful contribution addressing AI safety through structured human oversight, though HITL concepts exist broadly\",\n \"m\": 3,\n \"mn\": \"Defines protocol concept and integration approach but lacks detailed specification or implementation examples\",\n \"o\": 2,\n \"on\": \"Minor similarities to general AI safety approaches but unique focus on signed confirmation objects\",\n \"mo\": 2,\n \"mon\": \"Single draft submission with no indication of working group interest or community adoption\",\n \"r\": 5,\n \"rn\": \"Directly addresses core AI agent safety and human-agent interaction concerns with practical protocol approach\",\n \"c\": [\"AI safety/alignment\", \"Human-agent interaction\", \"A2A protocols\"]\n}", "in_tok": 656, "out_tok": 250} {"draft_name": "draft-morrison-agent-channel-fan-out", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft specifies an agent-channel frame format for fan-out message delivery between concurrent agentic sessions of identity-bound principals within organizational contexts. It defines transport envelopes with identity attribution and provenance blocks, using Server-Sent Events streams with subscription filtering for ephemeral message routing.\",\n \"n\": 4,\n \"nn\": \"Notable originality in combining identity-keyed fan-out delivery with agentic session management and organizational identity substrates\",\n \"m\": 3,\n \"mn\": \"Defines protocol mechanisms with frame format, scope-expansion grammar, and delivery semantics but lacks implementation details\",\n \"o\": 2,\n \"on\": \"References related drafts but appears to introduce distinct concepts for agent-specific messaging patterns\",\n \"mo\": 2,\n \"mon\": \"Single revision with future date suggests early development stage without demonstrated community engagement\",\n \"r\": 5,\n \"rn\": \"Directly addresses core AI agent communication needs including identity management, concurrent sessions, and organizational structures\",\n \"c\": [\"A2A protocols\", \"Agent identity/auth\", \"Data formats/interop\"]\n}", "in_tok": 763, "out_tok": 266} {"draft_name": "draft-yakung-oauth-agent-attestation", "model": "claude-sonnet-4-20250514", "raw": "```json\n{\n \"s\": \"This draft defines ACAP, a cryptographic protocol for autonomous AI agent authentication using short-lived JWTs with delegatable, scope-limited permissions. The protocol includes delegation chains, audit logging, and human-in-the-loop approval mechanisms for agent credential management.\",\n \"n\": 4,\n \"nn\": \"Novel application of JWT delegation chains specifically for autonomous AI agents with tamper-evident audit trails and human instruction hashing\",\n \"m\": 4,\n \"mn\": \"Comprehensive 27-page specification covering credential format, delegation rules, verification algorithms, and audit structures\",\n \"o\": 2,\n \"on\": \"While using standard JWT/OAuth concepts, the AI agent-specific delegation chains and instruction hashing are relatively unique\",\n \"mo\": 2,\n \"mon\": \"Single draft from individual contributor with no apparent working group adoption or community feedback\",\n \"r\": 5,\n \"rn\": \"Directly addresses core AI agent authentication and authorization challenges with autonomous delegation capabilities\",\n \"c\": [\"Agent identity/auth\", \"AI safety/alignment\", \"Human-agent interaction\"]\n}\n```", "in_tok": 650, "out_tok": 265} {"draft_name": "draft-sharif-mcps-secure-mcp", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"MCPS adds a cryptographic security layer to the Model Context Protocol through agent passports, message signing, and replay protection while maintaining backward compatibility. The specification provides trust levels L0-L4 for progressive security negotiation between MCP endpoints.\",\n \"n\": 3,\n \"nn\": \"Solid security engineering applying established cryptographic primitives to MCP, with useful trust level gradation concept\",\n \"m\": 4,\n \"mn\": \"Detailed 43-page specification with specific cryptographic choices, encoding formats, and backward compatibility design\",\n \"o\": 2,\n \"on\": \"While security layers exist for other protocols, this is specifically tailored to MCP's JSON-RPC structure and agent interaction patterns\",\n \"mo\": 3,\n \"mon\": \"Active development indicated by recent March 2026 date, though unclear if this has working group adoption yet\",\n \"r\": 4,\n \"rn\": \"Directly addresses critical security needs for AI agent protocols, essential for production MCP deployments\",\n \"c\": [\"A2A protocols\", \"Agent identity/auth\", \"AI safety/alignment\"]\n}", "in_tok": 812, "out_tok": 268} {"draft_name": "draft-narvaneni-agent-uri", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"Defines agent:// URI scheme for addressing and discovering software agents across different systems. Provides layered conformance levels from basic addressing to full discovery with authentication.\",\n \"n\": 3,\n \"nn\": \"URI-based agent addressing is logical but not groundbreaking - extends familiar web concepts to agent domain\",\n \"m\": 4,\n \"mn\": \"50 pages suggests comprehensive specification with detailed protocol mechanics and layered conformance approach\",\n \"o\": 2,\n \"on\": \"Complements rather than duplicates existing agent protocols by focusing on addressing/discovery layer\",\n \"mo\": 2,\n \"mon\": \"Single draft from 2026 with no clear working group adoption signals\",\n \"r\": 5,\n \"rn\": \"Directly addresses core agent infrastructure need for standardized addressing and discovery mechanisms\",\n \"c\": [\"Agent discovery/reg\", \"Agent identity/auth\", \"A2A protocols\"]\n}", "in_tok": 620, "out_tok": 225} {"draft_name": "draft-templin-intarea-aero2", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft specifies AERO, an IP internetworking service for overlay multilink networks using IPv6 Neighbor Discovery for control plane messaging. It supports route optimization, mobility management, and multicast forwarding across air/land/sea/space networks with both provider-aggregated and provider-independent addressing.\",\n \"n\": 3,\n \"nn\": \"Useful contribution combining existing IPv6 ND with overlay multilink concepts for mobility applications\",\n \"m\": 4,\n \"mn\": \"Detailed 113-page specification with comprehensive protocol mechanisms and examples\",\n \"o\": 3,\n \"on\": \"Shares overlay networking and mobility concepts with existing drafts but has distinct OMNI interface approach\",\n \"mo\": 2,\n \"mon\": \"Single author draft with limited apparent working group engagement or community discussion\",\n \"r\": 1,\n \"rn\": \"Traditional networking protocol focused on route optimization and mobility - not related to AI or autonomous agents\",\n \"c\": [\"Other\"]\n}", "in_tok": 670, "out_tok": 240} {"draft_name": "draft-zw-opsawg-mcp-network-mgmt", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft extends the Model Context Protocol (MCP) to enable network equipment to act as MCP servers, allowing AI applications to interact with routers and switches for network management. It defines new capability tokens, tools, resources, and error codes specifically for network operations while maintaining backward compatibility.\",\"n\":2,\"nn\":\"Incremental extension applying existing MCP framework to network domain\",\"m\":3,\"mn\":\"Defines protocol extensions with clear mechanisms but lacks implementation details\",\"o\":3,\"on\":\"Shares JSON-RPC and MCP concepts with related drafts but focuses on network-specific domain\",\"mo\":2,\"mon\":\"Single revision with future date, no evidence of community engagement yet\",\"r\":4,\"rn\":\"Directly relevant as AI-driven network management using agent communication protocols\",\"c\":[\"Autonomous netops\",\"A2A protocols\",\"Data formats/interop\"]}", "in_tok": 553, "out_tok": 190} {"draft_name": "draft-josefsson-chempat", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft specifies Chempat, a generic family of Post-Quantum/Traditional hybrid Key Encapsulation Mechanisms that combines classical Diffie-Hellman key agreement with post-quantum cryptographic methods. It provides concrete instantiated algorithms combining traditional curves (P-256, X25519, etc.) with PQ methods like ML-KEM and NTRU Prime for protocol integration.\",\n \"n\": 2,\n \"nn\": \"Incremental work combining existing PQ and traditional cryptographic methods using standard hybrid approaches\",\n \"m\": 4,\n \"mn\": \"Detailed specification with concrete algorithm instances and specific curve/PQ method combinations ready for implementation\",\n \"o\": 4,\n \"on\": \"Significant overlap with existing hybrid PQ cryptography work and other IETF efforts on post-quantum key exchange\",\n \"mo\": 3,\n \"mon\": \"Active development in post-quantum cryptography space but unclear specific community momentum for this particular draft\",\n \"r\": 1,\n \"rn\": \"This is about post-quantum cryptography and key exchange mechanisms, not AI or agents - appears to be a false positive\",\n \"c\": [\"Other AI/agent\"]\n}", "in_tok": 651, "out_tok": 285} {"draft_name": "draft-yc-ipv6-for-ioa", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft analyzes how IPv6's address space and connectivity features can support a future Internet of Agents (IoA) with hundreds of billions of intelligent agents. It examines current IPv6 capabilities for agent infrastructure and identifies evolutionary requirements for IPv6 to better serve agent-to-agent communication.\",\n \"n\": 2,\n \"nn\": \"Incremental analysis applying existing IPv6 features to agent scenarios without introducing novel protocols or mechanisms\",\n \"m\": 2,\n \"mn\": \"Early sketch that identifies requirements and capabilities but lacks concrete protocol specifications or implementation details\",\n \"o\": 2,\n \"on\": \"Minor overlap with general IoT and IPv6 enhancement drafts, but IoA focus is relatively distinct\",\n \"mo\": 2,\n \"mn\": \"Single draft from 2026 with no indication of working group interest or community engagement\",\n \"r\": 4,\n \"rn\": \"Directly addresses AI agent infrastructure and agent-to-agent connectivity requirements at internet scale\",\n \"c\": [\"A2A protocols\", \"Agent discovery/reg\", \"Autonomous netops\"]\n}", "in_tok": 598, "out_tok": 262} {"draft_name": "draft-helixar-hdp-agentic-delegation", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"Specifies a cryptographic protocol for tracking delegation chains when AI agents act on behalf of humans, using signed tokens to create verifiable provenance records. The protocol enables offline verification of who authorized an agent and through what chain of delegation, without requiring central registries or third-party trust anchors.\",\n \"n\": 4,\n \"nn\": \"Novel approach to a genuinely new problem - cryptographic provenance for AI agent delegation chains is an emerging need with no established solutions\",\n \"m\": 4,\n \"mn\": \"Detailed protocol specification with Ed25519 signatures, token format, and verification procedures, appears implementation-ready\",\n \"o\": 1,\n \"on\": \"Unique approach addressing a specific gap in AI agent accountability that isn't covered by existing identity or authentication protocols\",\n \"mo\": 2,\n \"mon\": \"Single revision from 2026, appears to be early stage without evident working group adoption or community feedback\",\n \"r\": 5,\n \"rn\": \"Directly addresses core AI agent governance - delegation provenance is fundamental to responsible agentic AI deployment\",\n \"c\": [\"Agent identity/auth\", \"Policy/governance\", \"Human-agent interaction\"]\n}", "in_tok": 710, "out_tok": 280} {"draft_name": "draft-yue-anima-agent-recovery-networks", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft proposes a multi-agent framework for autonomous fault recovery in converged 5G/6G mobile networks, enabling distributed failure detection, diagnosis, and policy reconfiguration across heterogeneous network domains. It specifically targets complex deployments like Multi-Operator Core Networks where traditional centralized management approaches are insufficient.\",\n \"n\": 4,\n \"nn\": \"Novel application of multi-agent systems to mobile network fault recovery with task-oriented coordination and intent-driven policy reconfiguration across converged public-private domains\",\n \"m\": 3,\n \"mn\": \"Defines framework architecture and protocol requirements for inter-agent communication and state consistency, but lacks detailed implementation specifications and test scenarios\",\n \"o\": 2,\n \"on\": \"While autonomous network operations exist, the specific multi-agent approach for converged mobile network recovery appears relatively unique with minor conceptual overlap\",\n \"mo\": 2,\n \"mon\": \"Single draft submission with no clear indication of working group adoption or broader community engagement in this specific approach\",\n \"r\": 4,\n \"rn\": \"Directly relevant as it focuses on autonomous agent-based systems for network operations with explicit multi-agent collaboration and coordination mechanisms\",\n \"c\": [\"Autonomous netops\", \"A2A protocols\", \"Policy/governance\"]\n}", "in_tok": 639, "out_tok": 301} {"draft_name": "draft-morrison-mcp-dns-discovery", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft defines DNS TXT records for discovering Model Context Protocol (MCP) servers, including organizational identity and cryptographic binding for individual handles. It provides a DNS-based alternative to HTTPS-based discovery methods with DNSSEC validation requirements.\",\n \"n\": 3,\n \"nn\": \"Useful contribution applying established DNS discovery patterns to MCP servers with added cryptographic identity binding\",\n \"m\": 4,\n \"mn\": \"Detailed specification with three defined TXT record types, DNSSEC requirements, DANE pinning, and companion URI scheme\",\n \"o\": 3,\n \"on\": \"Shares DNS discovery concepts with existing RFCs (DKIM, SPF, DMARC) but applies them to novel MCP context\",\n \"mo\": 2,\n \"mn\": \"Single revision with future date suggests early development stage without apparent working group adoption\",\n \"r\": 4,\n \"rn\": \"Directly relevant to AI agent infrastructure by enabling discovery of Model Context Protocol servers\",\n \"c\": [\"Agent discovery/reg\", \"Agent identity/auth\", \"A2A protocols\"]\n}", "in_tok": 943, "out_tok": 266} {"draft_name": "draft-mozleywilliams-dnsop-dnsaid", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft proposes DNS-AID, a mechanism for AI agents to discover each other using DNS records within existing namespaces. It defines a structured approach for organizations to publish AI agent metadata and capabilities without requiring changes to DNS protocol itself.\",\n \"n\": 3,\n \"nn\": \"Applying DNS service discovery patterns to AI agents is a logical extension of existing techniques, but represents a useful contribution to AI infrastructure\",\n \"m\": 4,\n \"mn\": \"At 24 pages with structured DNS namespace definitions and clear protocol mechanisms, this appears to be a detailed specification ready for implementation\",\n \"o\": 3,\n \"on\": \"Shares service discovery concepts with existing DNS-SD work but applies them specifically to AI agent discovery domain\",\n \"mo\": 2,\n \"mon\": \"Single draft with February 2026 date suggests early stage with limited community engagement so far\",\n \"r\": 5,\n \"rn\": \"Directly addresses core AI agent infrastructure need for scalable discovery and interoperability between autonomous agents\",\n \"c\": [\"Agent discovery/reg\", \"A2A protocols\", \"Data formats/interop\"]\n}", "in_tok": 655, "out_tok": 268} {"draft_name": "draft-chen-oauth-rar-agent-extensions", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft extends OAuth 2.0 Rich Authorization Requests with three new capabilities: intent-based authorization for goal-oriented requests, policy context for specifying security assurance levels, and lifecycle binding to tie authorization validity to external process states. The extensions specifically target modern automated systems and autonomous agents that require more sophisticated authorization semantics than traditional OAuth flows.\",\n \"n\": 4,\n \"nn\": \"Notable originality in extending OAuth RAR with intent semantics and lifecycle binding - moves beyond traditional resource-based auth to goal-oriented authorization patterns suited for autonomous systems\",\n \"m\": 3,\n \"mn\": \"Defines specific protocol mechanisms and parameters but appears to lack detailed examples and implementation guidance that would make it immediately actionable\",\n \"o\": 2,\n \"on\": \"While building on OAuth RAR foundation, the specific extensions for intent, policy context, and lifecycle binding represent relatively unique approaches in the authorization space\",\n \"mo\": 2,\n \"mon\": \"Single revision with future date suggests early-stage work without established working group adoption or demonstrated community interest\",\n \"r\": 5,\n \"rn\": \"Directly addresses core challenges in agent authorization including goal-oriented permissions, policy-aware decisions, and lifecycle management - fundamental for autonomous agent ecosystems\",\n \"c\": [\"Agent identity/auth\", \"Policy/governance\", \"A2A protocols\"]\n}", "in_tok": 699, "out_tok": 314} {"draft_name": "draft-nandakumar-agent-sd-jwt", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft defines SD-Card, a selective disclosure JWT format for agent-to-agent systems that enables privacy-preserving discovery and identity management. It allows agents to selectively reveal capabilities and metadata while maintaining cryptographic integrity and preventing correlation across interactions.\",\n \"n\": 4,\n \"nn\": \"Notable integration of SD-JWT with agent systems for privacy-preserving discovery - addresses real correlation and privacy concerns in A2A interactions\",\n \"m\": 4,\n \"mn\": \"30 pages suggests detailed specification with SD-Card format definition, likely includes protocol mechanics and examples for implementation\",\n \"o\": 2,\n \"on\": \"While using established SD-JWT concepts, the application to agent discovery and A2A systems appears to be a relatively unique approach\",\n \"mo\": 2,\n \"mon\": \"Single draft from 2026 with no clear working group context - appears to be individual contribution without demonstrated community engagement\",\n \"r\": 5,\n \"rn\": \"Directly addresses core agent infrastructure needs - discovery, identity management, and privacy in A2A systems are fundamental requirements\",\n \"c\": [\"Agent discovery/reg\", \"Agent identity/auth\", \"A2A protocols\", \"Data formats/interop\"]\n}", "in_tok": 559, "out_tok": 288} {"draft_name": "draft-sharif-attp", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"ATTP defines a comprehensive trust framework for AI agents with 5 progressive trust levels, behavioral scoring, cryptographic identity, and real-time compliance gating. The protocol is transport-agnostic with bindings for MCP, REST, gRPC and other protocols to enable trust-based action authorization.\",\n \"n\": 4,\n \"nn\": \"Novel separation of trust from identity with multi-dimensional behavioral scoring and progressive trust levels specifically designed for autonomous agents\",\n \"m\": 4,\n \"mn\": \"Detailed specification with defined trust levels, scoring dimensions, cryptographic mechanisms, and multiple protocol bindings\",\n \"o\": 2,\n \"on\": \"Some conceptual overlap with general identity/auth frameworks but unique focus on agent trust scoring and behavioral limits\",\n \"mo\": 2,\n \"mn\": \"Single draft with no clear working group adoption or community feedback yet\",\n \"r\": 5,\n \"rn\": \"Directly addresses core challenge of autonomous agent authorization and trust management in multi-agent systems\",\n \"c\": [\"Agent identity/auth\", \"AI safety/alignment\", \"A2A protocols\", \"Policy/governance\"]\n}", "in_tok": 776, "out_tok": 271} {"draft_name": "draft-liu-nmrg-ai-llm-inference-requirements", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft analyzes requirements for large language model inference services by examining mainstream frameworks like vLLM and SGLang to understand their network and system demands. It aims to provide foundation work for defining unified LLM inference architectures to address scalability and deployment challenges.\",\"n\":2,\"nn\":\"Incremental survey work analyzing existing LLM inference frameworks without introducing novel technical approaches\",\"m\":2,\"mn\":\"Early requirements analysis with framework comparisons but lacks detailed technical specifications or concrete protocol definitions\",\"o\":3,\"on\":\"Shares conceptual ground with other AI infrastructure and model serving work but takes specific angle on inference requirements\",\"mo\":2,\"mon\":\"Single revision draft with no clear working group adoption or broad community engagement yet\",\"r\":5,\"rn\":\"Directly addresses core AI infrastructure topic of LLM inference serving which is fundamental to AI agent deployment\",\"c\":[\"Model serving/inference\",\"ML traffic mgmt\"]}", "in_tok": 651, "out_tok": 199} {"draft_name": "draft-dembowski-agentledger-proof-of-behavior", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"Defines a Proof-of-Behavior protocol for creating tamper-evident audit trails and pre-execution policy enforcement in autonomous AI agent systems. Uses signed receipts, hash-chains, and policy gate contracts to enable third-party verification that agents acted within declared behavioral rules.\",\n \"n\": 4,\n \"nn\": \"Novel application of cryptographic audit mechanisms specifically to AI agent behavioral verification - addresses emerging trust problem in autonomous systems\",\n \"m\": 3,\n \"mn\": \"Defines core protocol components (receipt format, hash-chain linking, policy gates) but lacks detailed specification examples and implementation guidance\",\n \"o\": 2,\n \"on\": \"Some conceptual overlap with general audit trail mechanisms but specific focus on AI agent pre-execution policy enforcement is relatively unique\",\n \"mo\": 2,\n \"mn\": \"Single revision draft with no apparent working group adoption or community feedback yet\",\n \"r\": 5,\n \"rn\": \"Directly addresses core AI agent challenges around trust, verification, and accountability - highly relevant to autonomous agent deployment\",\n \"c\": [\"AI safety/alignment\", \"Policy/governance\", \"Agent identity/auth\"]\n}", "in_tok": 621, "out_tok": 271} {"draft_name": "draft-guy-bary-stamp-protocol", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"STAMP introduces cryptographically verifiable delegation and proof tokens for AI agent systems, binding access to specific tasks and identities. The protocol extends OAuth 2.0/GNAP with task-bound tokens and message-level proofs for multi-agent environments following zero-trust principles.\",\"n\":4,\"nn\":\"Novel approach to agent-specific authorization with task-bound tokens and cryptographic proofs for multi-agent delegation - addresses emerging security needs not covered by traditional auth\",\"m\":2,\"mn\":\"Early sketch level - introduces concepts and framework but lacks detailed protocol specification, message formats, or implementation guidance\",\"o\":2,\"on\":\"Some conceptual overlap with OAuth extensions and PoP mechanisms but applies them uniquely to agent delegation scenarios\",\"mo\":2,\"mon\":\"Single revision draft with no clear working group adoption or community momentum visible\",\"r\":5,\"rn\":\"Core AI agent topic addressing critical security and authorization challenges in multi-agent systems with autonomous delegation\",\"c\":[\"Agent identity/auth\",\"A2A protocols\",\"AI safety/alignment\"]}", "in_tok": 611, "out_tok": 226} {"draft_name": "draft-sharif-apki-agent-pki", "model": "claude-sonnet-4-20250514", "raw": "```json\n{\n \"s\": \"This draft proposes Agent Public Key Infrastructure (APKI), extending X.509 certificates with agent-specific capabilities like trust scoring, delegation chains, and model provenance for AI agent identity. It includes agent:// URI schemes, transparency logs, and cross-organizational federation while maintaining compatibility with existing PKI systems.\",\n \"n\": 4,\n \"nn\": \"Notable originality in adapting PKI specifically for AI agents with novel extensions for trust scoring, capability constraints, and model provenance - goes well beyond simple X.509 extension\",\n \"m\": 3,\n \"mn\": \"Defines concrete protocol mechanisms including URI schemes, certificate extensions, and transparency logs, but lacks detailed implementation examples and test vectors\",\n \"o\": 2,\n \"on\": \"Some conceptual overlap with general PKI and identity systems, but the AI agent-specific adaptations and extensions are fairly unique\",\n \"mo\": 2,\n \"mon\": \"Future-dated draft (2026) suggests early development stage, unclear community engagement or working group adoption at this point\",\n \"r\": 5,\n \"rn\": \"Core AI agent infrastructure addressing fundamental identity and trust challenges for autonomous agents performing authenticated actions\",\n \"c\": [\"Agent identity/auth\", \"A2A protocols\", \"Policy/governance\"]\n}\n```", "in_tok": 713, "out_tok": 304} {"draft_name": "draft-liao-aipref-autoctl-core", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft specifies a machine-readable protocol for controlling AI-driven web automation permissions, extending RFC9309 to cover broader state-changing activities. It defines file formats, HTTP method restrictions, and purpose requirements for server-side automation control.\",\n \"n\": 3,\n \"nn\": \"Useful contribution extending existing work to address AI automation control needs\",\n \"m\": 2,\n \"mn\": \"Early sketch stage - lacks detailed protocol specification and implementation examples\",\n \"o\": 3,\n \"on\": \"Builds on RFC9309 and shares concepts with web automation control mechanisms\",\n \"mo\": 2,\n \"mon\": \"Single revision with no apparent working group activity or community engagement\",\n \"r\": 4,\n \"rn\": \"Directly addresses AI agent automation control and permissions - highly relevant\",\n \"c\": [\"Policy/governance\", \"Agent identity/auth\", \"Data formats/interop\"]\n}", "in_tok": 523, "out_tok": 222} {"draft_name": "draft-ietf-anima-brski-ae", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This IETF draft extends the BRSKI protocol to support alternative certificate enrollment mechanisms beyond EST, such as CMP, using authenticated self-contained signed objects. It addresses deployment scenarios where standard EST enrollment may not be optimal and updates the BRSKI architecture accordingly.\",\"n\":2,\"nn\":\"Incremental extension of existing BRSKI protocol to support additional enrollment methods - useful but not particularly novel\",\"m\":4,\"mn\":\"Detailed specification with clear protocol extensions, architecture updates, and specific enrollment mechanisms defined\",\"o\":3,\"on\":\"Shares concepts with existing BRSKI and certificate enrollment drafts but provides distinct alternative approach\",\"mo\":4,\"mon\":\"Active IETF working group draft with clear progression and community interest in ANIMA WG\",\"r\":1,\"rn\":\"This is about network device certificate enrollment and bootstrapping - not related to AI or autonomous agents despite 'AE' abbreviation\",\"c\":[\"Other AI/agent\"]}", "in_tok": 636, "out_tok": 208} {"draft_name": "draft-sharif-agent-audit-trail", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft specifies a JSON-based standard logging format for autonomous AI systems with tamper-evident hash chaining and compliance mappings to EU AI Act and other regulations. It addresses audit requirements through structured records covering agent identity, actions, outcomes, and trust levels with privacy-preserving features.\",\n \"n\": 3,\n \"nn\": \"Useful contribution applying established audit logging concepts to AI agents with regulatory compliance focus, though the core technical approach of hash-chained JSON logs is not novel\",\n \"m\": 4,\n \"mn\": \"Detailed specification with clear field definitions, hash chaining mechanism, multiple export formats, and concrete regulatory mappings - appears implementation-ready\",\n \"o\": 2,\n \"on\": \"Some conceptual overlap with general audit logging standards but specific focus on AI agent requirements and regulatory compliance creates distinct value\",\n \"mo\": 4,\n \"mn\": \"Strong regulatory drivers from EU AI Act August 2026 deadline and compliance with multiple established frameworks suggests significant implementation interest\",\n \"r\": 5,\n \"rn\": \"Directly addresses core AI agent governance and auditability requirements with practical focus on regulatory compliance for autonomous systems\",\n \"c\": [\"Data formats/interop\", \"Policy/governance\", \"AI safety/alignment\"]\n}", "in_tok": 701, "out_tok": 295} {"draft_name": "draft-zhang-rtgwg-ai-agents-troubleshooting", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft outlines use cases and requirements for communication protocols that enable troubleshooting agents to operate on network devices. It appears to focus on defining how AI-powered diagnostic agents would interact and communicate within network infrastructure for automated problem resolution.\",\"n\":3,\"nn\":\"Useful contribution addressing agent communication for network troubleshooting, though builds on established concepts of network automation and distributed agents\",\"m\":2,\"mn\":\"Early stage draft focusing primarily on use cases and high-level requirements rather than concrete protocol specifications\",\"o\":3,\"on\":\"Shares concepts with existing network automation and agent communication work, but specific focus on troubleshooting agents provides some differentiation\",\"mo\":2,\"mon\":\"Single revision draft with no clear indication of working group adoption or broader community engagement\",\"r\":4,\"rn\":\"Directly relevant to AI agents in networking context, specifically addressing agent-to-agent communication for automated network operations\",\"c\":[\"A2A protocols\",\"Autonomous netops\",\"Agent discovery/reg\"]}", "in_tok": 488, "out_tok": 214} {"draft_name": "draft-dunbar-neotec-ac-te-applicability", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft explores applying existing IETF YANG data models for Attachment Circuit services and Traffic Engineering to support dynamic AI model placement at edge cloud sites. The work focuses on selecting optimal edge locations for real-time AI inference based on network performance between street cameras and compute nodes.\",\n \"n\": 2,\n \"nn\": \"Incremental application of existing YANG models to edge AI placement - useful but not particularly novel\",\n \"m\": 3,\n \"mn\": \"Defines specific mechanisms for applying AC and TE models with concrete use case mapping, though likely needs more implementation details\",\n \"o\": 3,\n \"on\": \"Shares concepts with other edge computing and traffic engineering work, but specific AI placement angle provides some differentiation\",\n \"mo\": 2,\n \"mon\": \"Single revision with June 2025 date suggests early stage work without demonstrated community traction yet\",\n \"r\": 4,\n \"rn\": \"Directly relevant to AI infrastructure with focus on dynamic model placement and inference optimization at network edge\",\n \"c\": [\"Model serving/inference\", \"ML traffic mgmt\", \"Autonomous netops\"]\n}", "in_tok": 630, "out_tok": 268} {"draft_name": "draft-templin-6man-aero3", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft specifies AERO (Automatic Extended Route Optimization) for IP networking over OMNI interfaces, using IPv6 Neighbor Discovery for secure network admission and forwarding. It targets mobile internetworking applications across air/land/sea/space domains including aviation and intelligent transportation systems.\",\n \"n\": 2,\n \"nn\": \"Incremental extension of existing IPv6 ND and route optimization concepts to overlay multilink networks\",\n \"m\": 4,\n \"mn\": \"Detailed 99-page specification with comprehensive protocol mechanisms for mobility, multicast, and route optimization\",\n \"o\": 3,\n \"on\": \"Shares concepts with existing mobile IPv6 and overlay networking drafts but has distinct OMNI interface approach\",\n \"mo\": 2,\n \"mon\": \"Single author draft from 2026 with no clear working group adoption or community discussion\",\n \"r\": 1,\n \"rn\": \"Pure networking protocol for route optimization and mobility - no AI or agent-related content despite transportation applications\",\n \"c\": [\"Other AI/agent\"]\n}", "in_tok": 640, "out_tok": 259} {"draft_name": "draft-deforth-arp", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"ARP provides cryptographic verification of machine-readable entity claims published on domains using Ed25519 signatures and DNS TXT records, similar to DKIM but designed for AI agent consumption. The protocol aims to prevent narrative injection and entity spoofing in generative search systems by establishing domain-owner authorization of semantic data.\",\n \"n\": 3,\n \"nn\": \"Useful adaptation of established cryptographic DNS patterns (DKIM-like) to the specific problem of AI agent data verification, though the core technical approach is incremental\",\n \"m\": 4,\n \"mn\": \"Detailed specification with clear protocol mechanics, signature schemes, and DNS integration patterns suitable for implementation\",\n \"o\": 2,\n \"on\": \"Some conceptual similarity to existing DNS-based authentication mechanisms but addresses a distinct use case for AI agent data verification\",\n \"mo\": 2,\n \"mn\": \"Single revision with future date suggests early stage, unclear community engagement or working group involvement\",\n \"r\": 5,\n \"rn\": \"Directly addresses core AI agent challenges around data authenticity, trust, and preventing adversarial content injection in RAG systems\",\n \"c\": [\"Agent identity/auth\", \"AI safety/alignment\", \"Data formats/interop\"]\n}", "in_tok": 648, "out_tok": 293} {"draft_name": "draft-song-anp-aitp", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"AITP defines a message-framed transport protocol above AIP for AI agent communication, with native method awareness and invocation semantics. Unlike traditional byte-stream transports, it provides first-class support for agent method calls, request-response correlation, and streaming between named autonomous agents.\",\"n\":4,\"nn\":\"Novel agent-native transport design with method awareness as first-class protocol feature, moving beyond adapting existing protocols\",\"m\":4,\"mn\":\"Detailed 34-page specification with segment formats, state machines, flow control, and protocol interfaces\",\"o\":2,\"on\":\"Builds on AIP foundation but introduces distinct agent-centric transport concepts not found in traditional protocols\",\"mo\":2,\"mon\":\"Single draft submission with no clear working group adoption or community feedback\",\"r\":5,\"rn\":\"Core AI agent infrastructure addressing fundamental transport needs for autonomous agent communication\",\"c\":[\"A2A protocols\",\"Agent identity/auth\"]}", "in_tok": 882, "out_tok": 203} {"draft_name": "draft-sovereign-haip", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"HAIP proposes a real-time identity verification protocol for AI agents using microsecond pulses and decentralized identity mechanisms. The draft addresses trust and integrity challenges in autonomous AI environments with NIST-aligned non-repudiation.\",\n \"n\": 4,\n \"nn\": \"Novel combination of microsecond-level timing requirements with decentralized identity teleportation for AI agents represents genuinely creative approach\",\n \"m\": 2,\n \"mn\": \"Abstract-level description only with unusual technical claims but no actual protocol specification or implementation details\",\n \"o\": 2,\n \"on\": \"Some conceptual overlap with general identity protocols but the AI-specific microsecond pulse approach appears unique\",\n \"mo\": 1,\n \"mon\": \"Single draft from 2026 with no indication of working group adoption or community engagement\",\n \"r\": 5,\n \"rn\": \"Directly addresses core AI agent challenges of identity verification and autonomous system trust in networked environments\",\n \"c\": [\"Agent identity/auth\", \"AI safety/alignment\", \"A2A protocols\"]\n}", "in_tok": 539, "out_tok": 260} {"draft_name": "draft-aevum-causal-intervention-record", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"PACR defines a wire format for recording verifiable causal intervention events performed by autonomous agents, using a six-tuple structure that includes thermodynamic costs and Pearl's do-calculus classification. The format creates a content-addressed DAG of causal events without requiring wall-clock timestamps, making it suitable for distributed multi-agent systems.\",\"n\":5,\"nn\":\"Genuinely novel approach combining causal inference theory with thermodynamic accounting and distributed systems concepts for agent action recording\",\"m\":3,\"mn\":\"Well-defined protocol structure with clear mathematical formulation, though lacks implementation details and test vectors\",\"o\":1,\"on\":\"Unique approach to causal event recording with no apparent overlap with existing IETF work\",\"mo\":2,\"mon\":\"Single draft with no evidence of working group adoption or community engagement\",\"r\":5,\"rn\":\"Core AI agent topic addressing fundamental needs for verifiable action recording and causal reasoning in autonomous systems\",\"c\":[\"A2A protocols\",\"Data formats/interop\",\"Agent identity/auth\"]}", "in_tok": 821, "out_tok": 225} {"draft_name": "draft-pioli-agent-discovery", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"ARDP is a lightweight protocol for registering and discovering autonomous software agents in distributed environments, providing stable identities, endpoint resolution, and capability advertisement. The protocol supports multiple transport mechanisms and includes security-first discovery with minimal presence signaling.\",\"n\":3,\"nn\":\"Useful contribution addressing agent discovery needs, though builds on established service discovery patterns\",\"m\":3,\"mn\":\"Defines core protocol mechanisms and capabilities but lacks detailed implementation specifics and test cases\",\"o\":3,\"on\":\"Shares service discovery concepts with existing protocols like DNS-SD and SRV records, but applies them specifically to agent environments\",\"mo\":2,\"mon\":\"Single revision draft with no clear working group adoption or community feedback visible\",\"r\":5,\"rn\":\"Directly addresses core AI agent infrastructure needs for registration, discovery, and multi-protocol capability negotiation\",\"c\":[\"Agent discovery/reg\",\"Agent identity/auth\",\"A2A protocols\"]}", "in_tok": 564, "out_tok": 200} {"draft_name": "draft-wiethuechter-drip-det-moc", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft standardizes DRIP Entity Tags (DETs) as identifiers for drone systems in air traffic management, enabling privacy through session IDs, accountability via authentication keys, and confidentiality through encryption key IDs. It focuses on trustworthy air domain awareness for unmanned aircraft systems in remote identification and traffic management scenarios.\",\"n\":2,\"nn\":\"Incremental extension of existing DRIP protocol to standardize entity tag usage patterns\",\"m\":3,\"mn\":\"Defines specific mechanisms and usage patterns for DETs with clear protocol specifications\",\"o\":4,\"on\":\"Significant overlap with existing DRIP and UAS identification standards and related aviation protocols\",\"mo\":2,\"mon\":\"Limited evidence of broader community engagement beyond single revision activity\",\"r\":2,\"rn\":\"Tangentially AI-related through autonomous drone systems but primarily focused on identification protocols rather than AI agents\",\"c\":[\"Agent identity/auth\",\"Policy/governance\"]}", "in_tok": 574, "out_tok": 202} {"draft_name": "draft-ni-wimse-ai-agent-identity", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft explores applying WIMSE (Workload Identity in Multi-System Environments) frameworks to establish independent identities and credential management for AI agents. It addresses the need for AI agents to have their own authentication and authorization mechanisms separate from their human operators or hosting systems.\",\"n\":3,\"nn\":\"Useful application of existing identity frameworks to AI agents, addressing a genuine need but building on established WIMSE concepts\",\"m\":2,\"mn\":\"Early stage applicability document with basic framework mapping, lacks detailed protocol specifications or implementation guidance\",\"o\":3,\"on\":\"Shares identity management concepts with other agent authentication drafts and builds directly on WIMSE specifications\",\"mo\":3,\"mon\":\"Active development as indicated by recent date, but momentum unclear without revision history or working group status\",\"r\":5,\"rn\":\"Directly addresses core AI agent identity and authentication challenges, highly relevant to autonomous agent deployments\",\"c\":[\"Agent identity/auth\"]}", "in_tok": 482, "out_tok": 203} {"draft_name": "draft-rehfeld-apix-core", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"Proposes APIX, a global service discovery infrastructure designed specifically for autonomous agents to find and consume APIs, using HATEOAS principles and a three-dimensional trust model. Defines core governance, commercial sustainability, and indexing mechanisms as foundational infrastructure for machine-native service discovery.\",\n \"n\": 4,\n \"nn\": \"Notable originality in proposing a comprehensive, commercially sustainable global infrastructure specifically designed for autonomous agents rather than adapting human-centric discovery mechanisms\",\n \"m\": 4,\n \"mn\": \"Detailed specification with governance model, trust framework, manifest format, commercial mechanisms, and API definitions across 61 pages with concrete implementation guidance\",\n \"o\": 2,\n \"on\": \"Minor similarities to service discovery protocols but unique in its autonomous agent focus and comprehensive commercial/governance framework\",\n \"mo\": 2,\n \"mon\": \"Single revision with no apparent working group adoption or community discussion visible\",\n \"r\": 5,\n \"rn\": \"Core AI agent infrastructure addressing fundamental service discovery needs for autonomous systems operating independently on the internet\",\n \"c\": [\"Agent discovery/reg\", \"Policy/governance\", \"A2A protocols\"]\n}", "in_tok": 721, "out_tok": 274} {"draft_name": "draft-bernardos-green-isac-uc", "model": "claude-sonnet-4-20250514", "raw": "```json\n{\n \"s\": \"This draft presents a use case for Integrated Sensing and Communications (ISAC) within the GREEN Working Group framework. It proposes a structured template for GREEN use cases while exploring how ISAC can enable environment perception and object tracking alongside data transmission.\",\n \"n\": 2,\n \"nn\": \"Incremental work applying existing ISAC concepts to GREEN framework - the sensing+communication integration is well-established, contribution is mainly in structuring it as a GREEN use case\",\n \"m\": 2,\n \"mn\": \"Early sketch that proposes a template and presents a use case but lacks detailed technical mechanisms, protocol definitions, or implementation specifics\",\n \"o\": 3,\n \"on\": \"Shares conceptual overlap with other GREEN working group drafts and general ISAC literature, but specific application to GREEN use case template appears reasonably distinct\",\n \"mo\": 3,\n \"mon\": \"Recent draft with active development indicated by 2025 date, appears to be part of ongoing GREEN WG activities but unclear level of broader community engagement\",\n \"r\": 3,\n \"rn\": \"Partially relevant - focuses on sensing and communications integration which touches on autonomous systems and smart environments, but not directly about AI agents or their specific networking needs\",\n \"c\": [\"Other AI/agent\"]\n}\n```", "in_tok": 644, "out_tok": 308} {"draft_name": "draft-shi-moq-kvcache", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft specifies how to transmit KVCache (key-value cache from LLM inference) over MoQT (Media over QUIC Transport) to enable caching and reuse across nodes. It addresses the challenge of efficiently transferring large KVCache data structures used in LLM prefill/decode phases.\",\"n\":3,\"nn\":\"Applies existing MoQT transport to a new AI use case - useful but incremental application\",\"m\":2,\"mn\":\"Early stage draft with basic problem statement and approach but lacks detailed protocol specification\",\"o\":2,\"on\":\"Relates to general MoQT work but KVCache application appears relatively unique\",\"mo\":2,\"mon\":\"Single revision from October 2025, too recent to assess ongoing development\",\"r\":4,\"rn\":\"Directly addresses AI model serving infrastructure and LLM inference optimization\",\"c\":[\"Model serving/inference\",\"ML traffic mgmt\"]}", "in_tok": 605, "out_tok": 205} {"draft_name": "draft-reilly-sentinel-protocol", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"Defines the Reilly Sentinel Protocol (RSP) for blockchain-anchored integrity and provenance tracking across the AI lifecycle, from datasets to inference outputs. Uses quantum-resistant triple-hash architecture with cross-chain proofs and DOI-based identity stabilization for evidence-grade audit trails.\",\n \"n\": 4,\n \"nn\": \"Notable integration of quantum-resistant cryptography with cross-chain blockchain anchoring specifically for AI provenance - combines multiple emerging areas in a novel way\",\n \"m\": 4,\n \"mn\": \"Detailed 25-page specification with formal protocol definition, multiple serialization formats, threat models, and implementation status - approaching implementation-ready\",\n \"o\": 2,\n \"on\": \"While blockchain and provenance concepts exist separately, the specific combination for AI lifecycle integrity with quantum resistance appears unique\",\n \"mo\": 3,\n \"mn\": \"Second revision (-01) with substantial expansions suggests active development, though no clear working group adoption indicated\",\n \"r\": 5,\n \"rn\": \"Directly addresses core AI governance and safety concerns around model provenance, training integrity, and audit trails - highly relevant to AI agent ecosystems\",\n \"c\": [\"AI safety/alignment\", \"Data formats/interop\", \"Policy/governance\", \"Agent identity/auth\"]\n}", "in_tok": 881, "out_tok": 306} {"draft_name": "draft-xu-rtgwg-fare-in-sun", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft adapts Weighted Equal-Cost Multi-Path (WECMP) load balancing, called Fully Adaptive Routing Ethernet (FARE), from scale-out to scale-up networks for AI workloads. It targets ultra-high-throughput, ultra-low-latency networking requirements for Mixture of Experts architectures across hundreds of GPUs.\",\"n\":2,\"nn\":\"Incremental extension of existing WECMP mechanisms to a different network topology, not fundamentally novel\",\"m\":2,\"mn\":\"Early stage draft with basic problem statement and high-level approach, lacks detailed protocol specification\",\"o\":3,\"on\":\"Shares load balancing concepts with existing multipath routing work, but specific AI scale-up network focus is somewhat distinct\",\"mo\":2,\"mon\":\"Single revision draft with no indication of working group adoption or broader community engagement\",\"r\":4,\"rn\":\"Directly addresses AI infrastructure networking needs for large-scale model training and inference\",\"c\":[\"ML traffic mgmt\"]}", "in_tok": 638, "out_tok": 224} {"draft_name": "draft-herbert-fast", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"FAST protocol enables hosts to request firewall traversal and network service tickets from local agents, with tickets carried in IPv6 Hop-by-Hop options to signal granted rights or service requests. The protocol provides a generic framework for network admission control and service provisioning through ticket-based signaling.\",\"n\":3,\"nn\":\"Useful contribution applying ticket-based access control to network services, though builds on established concepts\",\"m\":3,\"mn\":\"Defines protocol mechanics and ticket structure but lacks detailed implementation specifics\",\"o\":2,\"on\":\"Some conceptual overlap with network access control protocols but distinct ticket-based approach\",\"mo\":2,\"mon\":\"Appears to be early draft with limited revision history\",\"r\":2,\"rn\":\"Tangentially related to agents as local ticket-issuing entities, but primarily a network protocol\",\"c\":[\"Autonomous netops\"]}", "in_tok": 629, "out_tok": 190} {"draft_name": "draft-tong-network-agent-use-cases-in-6g", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft outlines use cases for network AI agents in 6G systems, focusing on connectivity services and third-party applications based on 3GPP scenarios. It discusses integration requirements and interaction workflows for AI agents within the 6G network framework.\",\n \"n\": 2,\n \"nn\": \"Incremental work that applies existing AI agent concepts to 6G networking scenarios without introducing fundamentally new approaches\",\n \"m\": 2,\n \"mn\": \"Early-stage draft that presents use cases and high-level requirements but lacks detailed technical specifications or concrete protocols\",\n \"o\": 4,\n \"on\": \"Significant overlap with existing 6G network management and AI-driven network operations literature, heavily based on 3GPP technical reports\",\n \"mo\": 3,\n \"mon\": \"Active development in 6G standardization space with moderate interest, though specific IETF momentum unclear from single snapshot\",\n \"r\": 4,\n \"rn\": \"Directly relevant to AI agents in networking context, specifically addressing autonomous network operations and AI-driven service management\",\n \"c\": [\"Autonomous netops\", \"ML traffic mgmt\", \"Policy/governance\", \"Agent discovery/reg\"]\n}", "in_tok": 560, "out_tok": 282} {"draft_name": "draft-dunbar-onsen-ac-te-applicability", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This document evaluates how existing IETF YANG data models for Attachment Circuit services and Traffic Engineering can support dynamic AI model placement at edge cloud sites. It focuses on optimizing real-time AI inference by selecting edge locations based on end-to-end network performance metrics between street cameras and compute nodes.\",\"n\":2,\"nn\":\"Incremental application of existing YANG models to edge AI placement - straightforward extension of current standards\",\"m\":3,\"mn\":\"Defines specific protocol application with network performance mapping but lacks detailed implementation specifications\",\"o\":3,\"on\":\"Shares Traffic Engineering and edge computing concepts with other IETF work, moderate conceptual overlap\",\"mo\":2,\"mon\":\"Single revision with future date suggests early stage development, limited community engagement visible\",\"r\":4,\"rn\":\"Directly addresses AI inference optimization in network infrastructure with practical edge deployment scenarios\",\"c\":[\"ML traffic mgmt\",\"Autonomous netops\",\"Model serving/inference\"]}", "in_tok": 629, "out_tok": 208} {"draft_name": "draft-jimenez-agent-directory", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft adapts the CoRE Resource Directory from IoT to software agents, creating an Agent Directory service for agent registration, capability advertisement, and discovery. It reuses the RFC 9176 registration and lookup model but applies it to software agents using HTTP and JSON instead of constrained devices.\",\"n\":2,\"nn\":\"Straightforward adaptation of existing IoT resource directory concepts to software agents with protocol translation\",\"m\":3,\"mn\":\"Defines a complete directory service protocol with registration, lookup, and lifecycle management mechanisms\",\"o\":4,\"on\":\"Heavy reuse of RFC 9176 CoRE Resource Directory concepts and mechanisms, essentially a protocol port\",\"mo\":2,\"mon\":\"Single revision with no clear working group adoption or broader community engagement\",\"r\":4,\"rn\":\"Directly addresses agent discovery and registration, a fundamental requirement for multi-agent systems\",\"c\":[\"Agent discovery/reg\",\"A2A protocols\"]}", "in_tok": 562, "out_tok": 201} {"draft_name": "draft-diaconu-agents-authz-info-sharing", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"Addresses authorization information sharing challenges in distributed multi-agent systems operating across multiple administrative domains with different identity providers. Proposes solutions using dynamic identity, interoperable claims, and verifiable credentials for secure cross-domain authorization.\",\"n\":3,\"nn\":\"Useful contribution applying existing identity/authorization concepts to multi-agent distributed systems context\",\"m\":2,\"mn\":\"Early sketch stage with conceptual discussion but lacking detailed protocol specifications\",\"o\":4,\"on\":\"Significant overlap with existing OAuth, federated identity, and verifiable credentials work\",\"mo\":2,\"mon\":\"Single revision with future date suggesting very early stage development\",\"r\":4,\"rn\":\"Directly relevant to AI agent systems, specifically addressing critical cross-domain authorization challenges\",\"c\":[\"Agent identity/auth\",\"A2A protocols\",\"Policy/governance\"]}", "in_tok": 538, "out_tok": 177} {"draft_name": "draft-lake-pocero-authkem-ikr-edhoc", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This IETF draft specifies an optimized KEM-based authentication method for EDHOC that enables post-quantum authentication in constrained environments where the initiator knows the responder's credentials in advance. The approach reduces the handshake to three mandatory messages while maintaining mutual authentication, forward secrecy, and identity protection using post-quantum cryptographic mechanisms like ML-KEM.\",\n \"n\": 3,\n \"nn\": \"Useful optimization of existing KEM-based EDHOC authentication for a specific but common scenario in IoT/constrained environments\",\n \"m\": 4,\n \"mn\": \"Detailed specification with concrete protocol mechanics, though implementation examples and test vectors not evident from abstract\",\n \"o\": 4,\n \"on\": \"Builds directly on existing KEM-based Authentication for EDHOC draft, representing a scenario-specific optimization rather than fundamentally new approach\",\n \"mo\": 3,\n \"mon\": \"Recent draft from January 2026 suggesting active development, but momentum unclear from single document snapshot\",\n \"r\": 2,\n \"rn\": \"Authentication protocol for constrained devices is only tangentially related to AI agents - primarily a cryptographic protocol optimization\",\n \"c\": [\"Agent identity/auth\"]\n}", "in_tok": 651, "out_tok": 295} {"draft_name": "draft-deforth-arp-reasoning-protocol", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"Defines Agentic Reasoning Protocol (ARP) v2.0 for entities to publish machine-readable context, corrections, and boundaries directly to AI agents and RAG systems. Extends static file approach with REST APIs, cryptographic attestation, DIDs, and real-time updates via Server-Sent Events.\",\n \"n\": 4,\n \"nn\": \"Novel approach to structured AI agent communication that goes beyond existing web conventions like robots.txt, specifically targeting reasoning systems rather than crawlers\",\n \"m\": 4,\n \"mn\": \"Detailed 29-page specification with REST API definitions, cryptographic components, DID integration, and comprehensive examples using proper RFC conventions\",\n \"o\": 2,\n \"on\": \"Some conceptual overlap with existing web standards (robots.txt, schema.org) but targets fundamentally different use case of reasoning agents vs content indexing\",\n \"mo\": 2,\n \"mn\": \"Single draft from 2026 with no apparent working group adoption or community feedback yet visible\",\n \"r\": 5,\n \"rn\": \"Directly addresses core challenge of how AI agents should interact with authoritative sources and obtain reliable context for reasoning tasks\",\n \"c\": [\"A2A protocols\", \"Agent identity/auth\", \"Data formats/interop\", \"Policy/governance\"]\n}", "in_tok": 741, "out_tok": 310} {"draft_name": "draft-han-rtgwg-codeployment-pfc-fgfc", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft proposes fine-grained flow control mechanisms for WANs that collaborate with existing PFC mechanisms in data centers to achieve end-to-end lossless transmission. It describes use cases and requirements for coordinating flow control across DCN and WAN boundaries.\",\n \"n\": 2,\n \"nn\": \"Incremental extension of existing flow control concepts from DCNs to WANs with coordination mechanisms\",\n \"m\": 2,\n \"mn\": \"Early problem statement and use case description without detailed protocol mechanisms or specifications\",\n \"o\": 3,\n \"on\": \"Shares flow control concepts with existing DCN work but addresses specific DCN-WAN coordination gap\",\n \"mo\": 2,\n \"mon\": \"Single revision draft from February 2026 with no apparent working group adoption yet\",\n \"r\": 1,\n \"rn\": \"Pure networking flow control draft with no AI, agent, or autonomous system components\",\n \"c\": [\"Other AI/agent\"]\n}", "in_tok": 646, "out_tok": 239} {"draft_name": "draft-drake-agent-identity-registry", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"Proposes a federated architecture for issuing hardware-anchored identities to AI agents and autonomous systems using TPMs and secure enclaves to prevent Sybil attacks. The system uses a three-tier DNS-like structure with governance authorities, registry operators, and registrars that issue standard OIDC tokens.\",\n \"n\": 4,\n \"nn\": \"Novel application of hardware attestation to autonomous agent identity with economic Sybil resistance through required physical devices\",\n \"m\": 4,\n \"mn\": \"Detailed 58-page specification with defined architecture, protocols, and URN namespace, though lacks implementation examples\",\n \"o\": 2,\n \"on\": \"Unique combination of hardware attestation and federated identity for agents, minimal overlap with existing identity systems\",\n \"mo\": 2,\n \"mon\": \"Individual draft from 2026 with no clear working group adoption or community discussion yet\",\n \"r\": 5,\n \"rn\": \"Directly addresses core challenge of autonomous agent identity verification and authentication in networked environments\",\n \"c\": [\"Agent identity/auth\", \"Policy/governance\", \"Agent discovery/reg\"]\n}", "in_tok": 897, "out_tok": 275} {"draft_name": "draft-kotecha-agentic-dispute-protocol", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft specifies the Agentic Dispute Protocol (ADP), a standardized framework for autonomous agents and AI systems to file, process, and resolve disputes through structured, automated processes. The protocol defines message formats, transport mechanisms, evidence submission standards, and cryptographic proof requirements for internet-native dispute resolution with support for multiple resolution frameworks.\",\n \"n\": 5,\n \"nn\": \"Genuinely novel approach - first standardized protocol specifically designed for autonomous agent dispute resolution with comprehensive technical and legal considerations\",\n \"m\": 4,\n \"mn\": \"Detailed specification with message formats, transport mechanisms, evidence standards, and cryptographic requirements - appears implementation-ready\",\n \"o\": 1,\n \"on\": \"Unique approach - no existing IETF work addresses automated dispute resolution for AI agents with this level of technical specification\",\n \"mo\": 2,\n \"mn\": \"Single revision draft with no indication of working group adoption or broader community involvement yet\",\n \"r\": 5,\n \"rn\": \"Core AI agent topic - directly addresses critical infrastructure need for autonomous agent interactions and automated contract enforcement\",\n \"c\": [\"A2A protocols\", \"Policy/governance\", \"Agent identity/auth\", \"Data formats/interop\"]\n}", "in_tok": 619, "out_tok": 290} {"draft_name": "draft-architect-cittamarket", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"Proposes CITTAMARKET Protocol for creating immutable AGI identity records using Bitcoin blockchain anchoring. Defines a four-layer architecture for unforgeable 'Sovereign Identity' records with cryptographic priority claims in specific Bitcoin blocks.\",\n \"n\": 2,\n \"nn\": \"Uses blockchain for identity anchoring - incremental application of existing concepts to AGI systems\",\n \"m\": 2,\n \"mn\": \"Early sketch at 3 pages - lacks detailed protocol specification, implementation details, or concrete examples\",\n \"o\": 3,\n \"on\": \"Shares blockchain-based identity concepts with existing work, though specific AGI focus is less common\",\n \"mo\": 1,\n \"mon\": \"Single draft with no apparent working group involvement or community engagement\",\n \"r\": 4,\n \"rn\": \"Directly addresses AGI identity management - a core concern for autonomous AI systems coordination\",\n \"c\": [\"Agent identity/auth\", \"Policy/governance\"]\n}", "in_tok": 588, "out_tok": 236} {"draft_name": "draft-drake-email-hardware-attestation", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"Proposes hardware attestation for email sender verification, enabling automated senders to prove access to genuine hardware security components via cryptographic evidence in message headers. Introduces persistent agent identities using URNs and privacy-preserving alternatives with SD-JWT for Sybil-resistant authentication.\",\"n\":4,\"nn\":\"Notable combination of hardware attestation with email authentication and agent identity management - creative approach to AI spam/abuse problem\",\"m\":4,\"mn\":\"Detailed specification with clear protocol mechanisms, header formats, and verification procedures though implementation details could be more complete\",\"o\":2,\"on\":\"Some conceptual overlap with existing email authentication (DKIM/SPF) but hardware attestation approach and agent identity focus is distinctive\",\"mo\":2,\"mon\":\"Single draft from 2026 with no apparent working group adoption or community discussion yet\",\"r\":5,\"rn\":\"Directly addresses critical AI agent authentication challenge as automated systems become prevalent in email communication\",\"c\":[\"Agent identity/auth\",\"A2A protocols\",\"AI safety/alignment\"]}", "in_tok": 868, "out_tok": 225} {"draft_name": "draft-hong-nmrg-ai-deploy", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft examines network and system configuration considerations for deploying AI inference services in distributed environments, covering client-cloud-edge architectures. It includes use cases like self-driving vehicles and network digital twins to illustrate deployment scenarios.\",\"n\":2,\"nn\":\"Incremental work applying existing distributed systems concepts to AI deployment without novel networking insights\",\"m\":2,\"mn\":\"Early sketch with general considerations but lacks concrete protocols, detailed mechanisms, or specific technical solutions\",\"o\":4,\"on\":\"Significant overlap with existing edge computing, distributed AI, and network optimization literature\",\"mo\":1,\"mon\":\"Single revision draft with no apparent working group activity or community engagement\",\"r\":4,\"rn\":\"Directly relevant to AI infrastructure deployment though focuses more on general distributed systems than AI-specific networking\",\"c\":[\"Model serving/inference\",\"ML traffic mgmt\"]}", "in_tok": 606, "out_tok": 183} {"draft_name": "draft-hood-independent-atp", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"Proposes Agent Transfer Protocol (ATP) as a dedicated application-layer protocol for AI agent traffic with agent-native methods like QUERY, DELEGATE, COLLABORATE and built-in identity mechanisms. Claims existing agent protocols only address messaging while HTTP lacks semantic vocabulary for agent systems at scale.\",\n \"n\": 4,\n \"nn\": \"Notable originality in proposing dedicated transport protocol for agents with semantic intent methods rather than building on HTTP\",\n \"m\": 3,\n \"mn\": \"Defines protocol structure and methods but lacks detailed specification and implementation guidance\",\n \"o\": 2,\n \"on\": \"Some conceptual overlap with agent communication protocols but unique transport-layer focus\",\n \"mo\": 2,\n \"mon\": \"Single revision from independent author without working group backing\",\n \"r\": 5,\n \"rn\": \"Core AI agent topic addressing fundamental transport and identity challenges for agentic systems\",\n \"c\": [\"A2A protocols\", \"Agent identity/auth\", \"ML traffic mgmt\"]\n}", "in_tok": 717, "out_tok": 241} {"draft_name": "draft-sarischo-6gip-aiagent-requirements", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft defines requirements for AI/ML agents in 6G networks, covering different agent types and their specific needs. It explains implementation approaches including API discovery and DNS-based agent discovery mechanisms for inter-domain and intra-domain scenarios.\",\n \"n\": 2,\n \"nn\": \"Incremental work applying existing AI agent concepts to 6G context without fundamentally new approaches\",\n \"m\": 2,\n \"mn\": \"Early sketch phase with basic requirements enumeration but lacks detailed protocol specifications or implementation details\",\n \"o\": 4,\n \"on\": \"Significant overlap with existing 3GPP AI/ML work and other 6G network intelligence drafts\",\n \"mo\": 2,\n \"mn\": \"Single revision with uncertain working group interest, limited community engagement evident\",\n \"r\": 4,\n \"rn\": \"Directly relevant to AI agents in network infrastructure with specific focus on discovery and requirements\",\n \"c\": [\"Agent discovery/reg\", \"Autonomous netops\", \"A2A protocols\"]\n}", "in_tok": 554, "out_tok": 245} {"draft_name": "draft-gudlab-agentid-protocol", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"AgentID defines an identity protocol for autonomous AI agents using signed JWT tokens (Agent Identity Tokens) that carry agent identity, owner verification, capabilities, and delegation chains. The protocol builds on OAuth 2.0/OpenID Connect to establish who controls an agent, what it can do, and accountability chains for agent actions across internet services.\",\n \"n\": 4,\n \"nn\": \"Notable originality in adapting OAuth/JWT for autonomous agents with delegation chains and capability attestation - addresses a genuine gap in existing identity protocols\",\n \"m\": 3,\n \"mn\": \"Defined protocol with clear token structure and verification flows, but lacks detailed message formats, error handling, and concrete implementation examples\",\n \"o\": 2,\n \"on\": \"Some conceptual overlap with OAuth/OpenID Connect adaptations, but the autonomous agent focus and delegation chains are relatively unique\",\n \"mo\": 2,\n \"mon\": \"Single revision with no clear working group adoption or broader community engagement visible\",\n \"r\": 5,\n \"rn\": \"Core AI agent topic addressing fundamental identity and accountability challenges for autonomous agents operating on the internet\",\n \"c\": [\"Agent identity/auth\", \"A2A protocols\", \"Policy/governance\"]\n}", "in_tok": 663, "out_tok": 286} {"draft_name": "draft-baur-pap", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"PAP defines a cryptographic protocol for agent-to-agent transactions controlled by human principals, using hierarchical delegation through signed mandates with selective disclosure and session ephemerality. The protocol requires no central registry or trusted third parties, building on existing cryptographic primitives.\",\n \"n\": 4,\n \"nn\": \"Notable originality in combining human-rooted trust model with agent delegation, context minimization at protocol level, and structural ephemerality guarantees\",\n \"m\": 4,\n \"mn\": \"Detailed 91-page specification suggests comprehensive protocol definition with implementation details, though actual test vectors unclear\",\n \"o\": 2,\n \"on\": \"Minor similarities to general delegation protocols but unique combination of human-agent trust model with selective disclosure\",\n \"mo\": 2,\n \"mon\": \"Single revision with future date suggests early stage, no evidence of working group adoption or community engagement\",\n \"r\": 5,\n \"rn\": \"Core AI agent topic addressing fundamental trust, delegation, and interaction protocols for autonomous agents\",\n \"c\": [\"A2A protocols\", \"Agent identity/auth\", \"Human-agent interaction\"]\n}", "in_tok": 553, "out_tok": 269} {"draft_name": "draft-smith-opsawg-ai-network-governance", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"Defines a comprehensive governance framework with thirteen areas (human authority, harm prevention, transparency, etc.) for AI systems that autonomously manage network devices. Addresses the critical gap in safety principles as AI moves from advisory to closed-loop autonomous operation in production networks.\",\n \"n\": 4,\n \"nn\": \"First comprehensive governance framework specifically for AI-mediated autonomous network management, addressing a critical gap as the industry moves toward closed-loop AI operations\",\n \"m\": 3,\n \"mn\": \"Well-structured framework with thirteen defined governance areas, but appears to lack detailed implementation guidance or specific technical mechanisms\",\n \"o\": 1,\n \"on\": \"Unique focus on governance framework for autonomous network device management - no other drafts address this specific intersection of AI governance and network operations\",\n \"mo\": 3,\n \"mon\": \"Recent draft (March 2026) addressing timely concerns about AI autonomy in critical infrastructure, though momentum unclear from single snapshot\",\n \"r\": 5,\n \"rn\": \"Core AI agent topic - directly addresses autonomous AI systems operating on network infrastructure with comprehensive governance considerations\",\n \"c\": [\"Policy/governance\", \"Autonomous netops\", \"AI safety/alignment\"]\n}", "in_tok": 654, "out_tok": 282} {"draft_name": "draft-mapmw-task-discovery", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft proposes an architectural framework for task discovery in agentic networks, where task owners publish structured task cards to platforms that enable autonomous agents to discover, negotiate, and execute tasks. The framework defines functional layers supporting secure task publication, agent discovery, capability evaluation, and multi-agent collaboration across heterogeneous systems.\",\"n\":4,\"nn\":\"Notable originality in proposing a task-centric discovery model that complements existing agent-to-agent approaches with structured task cards and bilateral negotiation mechanisms\",\"m\":3,\"mn\":\"Defines clear architectural framework with functional layers and interaction patterns, but appears to lack detailed protocol specifications and concrete implementation examples\",\"o\":2,\"on\":\"Minor similarities to existing agent discovery approaches like A2A and ARDP, but takes a distinct task-centric rather than agent-centric approach\",\"mo\":2,\"mon\":\"Single revision with future date (2026-03-24) suggests early stage work without established community momentum\",\"r\":5,\"rn\":\"Core AI agent topic directly addressing fundamental challenges in autonomous agent coordination, task discovery, and multi-agent collaboration\",\"c\":[\"Agent discovery/reg\",\"A2A protocols\",\"Data formats/interop\"]}", "in_tok": 777, "out_tok": 256} {"draft_name": "draft-li-dmsc-inf-architecture", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft proposes a network infrastructure architecture to support large-scale multi-agent collaboration, focusing on how network control and forwarding functions can actively participate in agent interactions. It addresses the infrastructure requirements and systematic approach for Dynamic Multi-agent Secured Collaboration (DMSC).\",\n \"n\": 4,\n \"nn\": \"Notable approach to infrastructure-level support for multi-agent systems, moving beyond application-layer solutions to network-aware collaboration\",\n \"m\": 2,\n \"mn\": \"Early architectural sketch without detailed protocol specifications or implementation mechanisms\",\n \"o\": 2,\n \"on\": \"Minor similarities to general network architecture work but unique focus on multi-agent collaboration infrastructure\",\n \"mo\": 2,\n \"mon\": \"Single revision from 2026 with no visible community engagement or working group activity\",\n \"r\": 4,\n \"rn\": \"Directly addresses AI agent networking and collaboration infrastructure, core to multi-agent systems deployment\",\n \"c\": [\"A2A protocols\", \"Autonomous netops\", \"Agent identity/auth\"]\n}", "in_tok": 534, "out_tok": 251} {"draft_name": "draft-anandakrishnan-ptv-attested-agent-identity", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"Specifies the Prove-Transform-Verify (PTV) protocol for hardware-anchored, zero-knowledge attested agent identity in federated environments. Uses TPM 2.0/secure enclaves, Groth16 ZK proofs, and HotStuff consensus for cross-domain agent authorization without exposing raw data.\",\n \"n\": 4,\n \"nn\": \"Novel combination of hardware attestation, zero-knowledge proofs, and jurisdiction-aware attestation chains for federated agent identity - addresses data gravity problem with creative technical approach\",\n \"m\": 3,\n \"mn\": \"Defined protocol with specific technical components (TPM 2.0, Groth16, HotStuff) and performance metrics, but lacks detailed implementation specifics and test vectors\",\n \"o\": 2,\n \"on\": \"While agent identity and attestation concepts exist, the specific combination of hardware anchoring with ZK proofs and jurisdiction-aware metadata appears relatively unique\",\n \"mo\": 2,\n \"mn\": \"Single draft from April 2026 with no indication of working group adoption or community feedback - appears to be individual submission\",\n \"r\": 5,\n \"rn\": \"Core AI agent topic directly addressing agent identity and authentication in federated environments - fundamental requirement for autonomous agent operations\",\n \"c\": [\"Agent identity/auth\", \"A2A protocols\", \"AI safety/alignment\"]\n}", "in_tok": 658, "out_tok": 332} {"draft_name": "draft-hillier-certisyn-ai-governance-verified", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft specifies a cryptographic verification standard for AI governance in regulated industries, defining verification objects, control areas, and maturity levels to produce auditor-grade attestations. It aims to create verifiable artifacts that complement existing governance frameworks like ISO/IEC 42001 and NIST AI RMF.\",\"n\":4,\"nn\":\"Notable originality in applying cryptographic attestation specifically to AI governance posture verification with deterministic reconstruction capabilities\",\"m\":3,\"mn\":\"Defines clear protocol structure with VRO objects, control areas, and maturity levels but lacks detailed implementation specifics\",\"o\":2,\"on\":\"Minor conceptual overlap with general governance frameworks but unique cryptographic verification approach for AI systems\",\"mo\":2,\"mon\":\"Single revision draft from 2026 with no apparent working group adoption or community activity\",\"r\":4,\"rn\":\"Directly relevant to AI governance with specific focus on agentic AI systems in regulated environments\",\"c\":[\"Policy/governance\",\"Agent identity/auth\",\"Data formats/interop\"]}", "in_tok": 651, "out_tok": 225} {"draft_name": "draft-ietf-rats-corim", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft specifies CBOR format for representing endorsements and reference values used in Remote Attestation Procedures (RATS). It defines the Concise Reference Integrity Manifest (CoRIM) to help verifiers assess the trustworthiness of remote attesters by providing standardized endorsement and reference value formats.\",\"n\":2,\"nn\":\"Incremental work applying existing CBOR encoding to RATS endorsement structures\",\"m\":4,\"mn\":\"Detailed 127-page specification with comprehensive information elements and likely includes examples\",\"o\":3,\"on\":\"Shares attestation concepts with other RATS drafts but focuses specifically on manifest format\",\"mo\":4,\"mon\":\"IETF working group draft with clear adoption and systematic naming convention\",\"r\":2,\"rn\":\"Tangentially related to AI agents - attestation could apply to AI systems but primarily about general device trustworthiness\",\"c\":[\"Agent identity/auth\"]}", "in_tok": 636, "out_tok": 206} {"draft_name": "draft-morrison-compute-location-gate", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft specifies a compute-location gate mechanism for negotiating where identity inference computations occur based on data provenance classes (active, passive aggregate, or passive individual). It enforces privacy constraints through wire-layer consent matching and refusal mechanisms, with passive individual observations staying local-only.\",\n \"n\": 4,\n \"nn\": \"Notable originality in combining provenance-based routing with wire-layer consent enforcement for identity inference, creating a novel privacy-preserving computation delegation framework\",\n \"m\": 3,\n \"mn\": \"Defines the core protocol mechanism and provenance classes but lacks detailed implementation specifics and examples\",\n \"o\": 2,\n \"on\": \"Minor conceptual overlap with privacy-preserving ML approaches but the provenance-class routing with wire-layer enforcement appears unique\",\n \"mo\": 2,\n \"mon\": \"Single revision draft with no clear working group activity or community engagement indicators\",\n \"r\": 4,\n \"rn\": \"Directly addresses AI agent identity inference with practical privacy controls relevant to agent-human interactions\",\n \"c\": [\"Agent identity/auth\", \"Policy/governance\", \"Model serving/inference\"]\n}", "in_tok": 773, "out_tok": 273} {"draft_name": "draft-tan-ccamp-uonaco-problem-statement", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft proposes unified orchestration between optical transport networks and AI computing infrastructure to optimize distributed AI workloads. It identifies the problem of isolated control planes preventing efficient coordination between network resources and compute schedulers in geographically distributed AI data centers.\",\n \"n\": 3,\n \"nn\": \"Useful contribution addressing real coordination gaps between optical networks and AI compute, though network-compute convergence concepts exist in other domains\",\n \"m\": 1,\n \"mn\": \"Pure problem statement draft outlining use cases and requirements without any protocol mechanisms or technical solutions\",\n \"o\": 2,\n \"on\": \"Minor conceptual overlap with general network-compute orchestration work but specifically focused on optical networks and AI workloads\",\n \"mo\": 2,\n \"mon\": \"Single revision with future date (2026-03-02) suggesting early-stage concept without demonstrated community engagement\",\n \"r\": 4,\n \"rn\": \"Directly relevant to AI infrastructure challenges, addressing critical network coordination needs for distributed AI training and inference\",\n \"c\": [\"ML traffic mgmt\", \"Autonomous netops\"]\n}", "in_tok": 743, "out_tok": 261} {"draft_name": "draft-ietf-ace-coap-est-oscore", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft specifies how to protect EST (Enrollment over Secure Transport) certificate provisioning using OSCORE instead of DTLS for constrained IoT environments. It builds on existing EST-CoAP specifications but substitutes OSCORE/EDHOC for transport security and optionally supports CBOR certificate formats.\",\"n\":2,\"nn\":\"Incremental work replacing DTLS with OSCORE in existing EST-CoAP framework\",\"m\":4,\"mn\":\"Detailed specification with clear protocol mechanics and security considerations\",\"o\":4,\"on\":\"Significant overlap with RFC9148 EST-coaps, mainly substituting security mechanisms\",\"mo\":4,\"mon\":\"IETF working group draft with established momentum in constrained device security\",\"r\":1,\"rn\":\"Pure certificate provisioning protocol for IoT devices, not related to AI or autonomous agents\",\"c\":[\"Other AI/agent\"]}", "in_tok": 634, "out_tok": 201} {"draft_name": "draft-sato-soos-hem", "model": "claude-sonnet-4-20250514", "raw": "```json\n{\n \"s\": \"Defines a kernel-level Human Escalation Mechanism (HEM) that puts AI agent sessions into a formal HEM_PENDING state when human oversight is required, routing structured escalation requests to human principals and enforcing decision protocols. The specification includes Policy Rationale Declarations (PRD) and Decision Rationale Records (DRR) to provide machine-readable reasoning for escalation policies and human decisions.\",\n \"n\": 4,\n \"nn\": \"Notable originality in treating human escalation as a first-class kernel session state rather than application-layer confirmation, with formal state transitions and structured decision types\",\n \"m\": 4,\n \"mn\": \"Detailed specification with clear state definitions, decision types, and integration with Cedar policies, though lacks implementation examples or test vectors\",\n \"o\": 2,\n \"on\": \"Minor conceptual overlap with human-in-the-loop systems but unique in its kernel-level approach and formal state management\",\n \"mo\": 3,\n \"mon\": \"Active development with clear regulatory motivation (EU AI Act Article 14) suggesting ongoing relevance and potential adoption pressure\",\n \"r\": 5,\n \"rn\": \"Directly addresses core challenge of human oversight in autonomous AI systems with technical specification for a fundamental interaction pattern\",\n \"c\": [\"Human-agent interaction\", \"Policy/governance\", \"AI safety/alignment\"]\n}\n```", "in_tok": 766, "out_tok": 325} {"draft_name": "draft-goswami-agentic-jwt", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft proposes Agentic JWT, an OAuth 2.0 extension for authenticating and authorizing autonomous AI agents through cryptographic agent identity and workflow-aware token binding. It addresses the 'intent-execution separation problem' where AI agents may deviate from user intent through dynamic workflow generation and autonomous decision-making.\",\n \"n\": 4,\n \"nn\": \"Notable originality in addressing agent identity through system prompt/tool checksums and workflow-aware token binding - genuinely novel approach to OAuth for AI agents\",\n \"m\": 4,\n \"mn\": \"Detailed 71-page specification with defined protocol mechanisms, new grant types, and PoP at agent level - appears implementation-ready though test vectors not mentioned\",\n \"o\": 2,\n \"on\": \"Minor similarities to general OAuth extensions and identity protocols, but the agentic-specific mechanisms appear quite unique\",\n \"mo\": 2,\n \"mn\": \"Single draft from 2026 with no apparent working group adoption or community feedback yet\",\n \"r\": 5,\n \"rn\": \"Core AI agent topic directly addressing fundamental authentication and authorization challenges in multi-agent systems\",\n \"c\": [\"Agent identity/auth\", \"A2A protocols\", \"AI safety/alignment\"]\n}", "in_tok": 806, "out_tok": 297} {"draft_name": "draft-papon-lake-pq-edhoc", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft extends EDHOC (Ephemeral Diffie-Hellman over COSE) to provide quantum resistance through two protocol variants that combine signatures and KEMs. The proposals aim to maintain lower message complexity compared to other quantum-resistant approaches while offering different tradeoffs between messages and computational overhead.\",\"n\":3,\"nn\":\"Useful contribution applying established post-quantum techniques to EDHOC, though the core approach of combining signatures with KEMs is not novel\",\"m\":4,\"mn\":\"Detailed 49-page specification with concrete protocol definitions for both 3-message and 3-or-4-message variants\",\"o\":4,\"on\":\"Significant overlap with other post-quantum EDHOC drafts, though claims to reduce message complexity compared to parallel efforts\",\"mo\":2,\"mon\":\"Single revision with no clear indication of working group adoption or broader community engagement\",\"r\":2,\"rn\":\"Primarily a cryptographic protocol for quantum-resistant key exchange, only tangentially related to AI/agents through general security applications\",\"c\":[\"Other AI/agent\"]}", "in_tok": 633, "out_tok": 237} {"draft_name": "draft-du-catalist-routing-considerations", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft proposes routing considerations for agentic networks where AI agents communicate with each other across domains using an overlay network architecture above IP. It addresses new networking requirements that emerge when agents need to connect to other agents with complementary skills during task execution.\",\"n\":3,\"nn\":\"Useful contribution addressing emerging agent-to-agent communication patterns, though overlay networking concepts are well-established\",\"m\":2,\"mn\":\"Early sketch at problem statement level - lacks detailed protocol mechanisms or concrete specifications\",\"o\":2,\"on\":\"Minor similarities to existing overlay networking approaches but focuses on novel agent-specific use case\",\"mo\":2,\"mon\":\"Single revision with limited evidence of broader community engagement or working group interest\",\"r\":4,\"rn\":\"Directly relevant to AI agent networking infrastructure and inter-agent communication protocols\",\"c\":[\"A2A protocols\",\"Agent discovery/reg\",\"Autonomous netops\"]}", "in_tok": 574, "out_tok": 191} {"draft_name": "draft-smith-rats-evidence-trans", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This IETF draft specifies methods for transforming different remote attestation evidence formats (DiceTcbInfo, concise evidence, SPDM measurements) into a common CoRIM internal representation. The goal is to enable Verifiers in Remote Attestation Procedures (RATS) to more easily appraise evidence from Attesters by standardizing the internal format while preserving original semantics.\",\"n\":2,\"nn\":\"Incremental work defining format transformations for existing attestation evidence types to a common representation\",\"m\":4,\"mn\":\"Detailed specification with concrete transformation methods for multiple evidence formats and appraisal procedures\",\"o\":3,\"on\":\"Shares concepts with existing RATS architecture and CoRIM drafts but provides distinct transformation mechanisms\",\"mo\":3,\"mon\":\"Active development in RATS working group addressing practical interoperability challenges\",\"r\":2,\"rn\":\"Tangentially related to AI agents as remote attestation could verify agent platforms, but primarily focused on general system trust verification\",\"c\":[\"Agent identity/auth\"]}", "in_tok": 613, "out_tok": 231} {"draft_name": "draft-wang-cats-odsi", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft proposes a framework for decentralized large language model inference across untrusted participants, treating inference as distributed layer-wise execution with explicit deadlines. The framework combines layer-aware transport, decentralized coordination, and security mechanisms for accountability without trusted execution environments.\",\n \"n\": 4,\n \"nn\": \"Notable originality in applying decentralized computing principles specifically to LLM inference with layer-aware routing and deadline-driven execution - addresses a genuine gap between centralized AI services and distributed computing\",\n \"m\": 2,\n \"mn\": \"Early architectural sketch with problem definition and solution space outlined, but explicitly states it does not specify concrete protocols, message formats, or state machines\",\n \"o\": 2,\n \"on\": \"Minor conceptual overlap with general distributed computing and federated learning approaches, but the specific focus on LLM layer-wise inference across untrusted networks is relatively unique\",\n \"mo\": 2,\n \"mon\": \"Single revision from 2026 with no clear working group adoption or community discussion visible, though the problem space is gaining attention\",\n \"r\": 5,\n \"rn\": \"Core AI topic directly addressing LLM inference infrastructure challenges with specific focus on network-level protocols and distributed execution mechanisms\",\n \"c\": [\"Model serving/inference\", \"A2A protocols\", \"Agent identity/auth\"]\n}", "in_tok": 788, "out_tok": 315} {"draft_name": "draft-campbell-agentic-http", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft proposes best practices for HTTP-based interactions between AI agents, including tool invocation via GET requests and formalizing the 402 status code for payment. It aims to standardize agentic HTTP interactions to prevent fragmented, proprietary ecosystems.\",\n \"n\": 3,\n \"nn\": \"Useful contribution applying existing HTTP semantics to agent interactions, though the core concepts are extensions of established patterns\",\n \"m\": 2,\n \"mn\": \"Early sketch with basic framework outlined but lacks detailed specifications, implementation guidance, and concrete examples\",\n \"o\": 2,\n \"on\": \"Some conceptual overlap with general HTTP best practices and web APIs, but specific focus on agentic interactions is relatively unique\",\n \"mo\": 2,\n \"mn\": \"Appears to be single revision with no clear working group adoption or broader community engagement yet\",\n \"r\": 5,\n \"rn\": \"Directly addresses core AI agent infrastructure needs for standardized communication protocols and interoperability\",\n \"c\": [\"A2A protocols\", \"Data formats/interop\", \"Agent identity/auth\"]\n}", "in_tok": 637, "out_tok": 263} {"draft_name": "draft-jiang-cats-reference-acn", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft proposes a reference model for AI-Agent Communication Networks (ACN) that extends the CATS framework to enable communication between AI agents. It defines reference points, protocol stacks, service provisioning models, and signaling procedures for AI agent networking.\",\n \"n\": 3,\n \"nn\": \"Useful application of CATS concepts to AI agent communication, though the core networking concepts are extensions of existing work\",\n \"m\": 2,\n \"mn\": \"Early sketch with basic model definition but lacks detailed protocol specifications and implementation details\",\n \"o\": 3,\n \"on\": \"Builds directly on existing CATS framework and shares concepts with other agent communication proposals\",\n \"mo\": 2,\n \"mon\": \"Recent draft with limited revision history and unclear working group engagement\",\n \"r\": 4,\n \"rn\": \"Directly addresses AI agent networking and communication infrastructure, which is highly relevant to AI agent systems\",\n \"c\": [\"A2A protocols\", \"Agent discovery/reg\", \"Data formats/interop\"]\n}", "in_tok": 576, "out_tok": 246} {"draft_name": "draft-gaikwad-llm-benchmarking-methodology", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft defines comprehensive benchmarking methodologies for evaluating LLM inference serving systems, specifying test procedures, measurement protocols, and reporting formats for performance characteristics. It serves as a companion to terminology standards and provides standardized approaches for measuring latency, throughput, and resource management in LLM deployments.\",\"n\":3,\"nn\":\"Useful contribution applying established benchmarking principles to the specific domain of LLM serving systems\",\"m\":4,\"mn\":\"Detailed specification with 48 pages covering test procedures, measurement specs, and reporting formats\",\"o\":2,\"on\":\"Some conceptual overlap with general performance benchmarking standards but focused specifically on LLM serving\",\"mo\":3,\"mon\":\"Active development as evidenced by 2026 date and companion terminology document reference\",\"r\":4,\"rn\":\"Directly relevant to AI systems focusing on standardized evaluation of model serving infrastructure\",\"c\":[\"Model serving/inference\"]}", "in_tok": 549, "out_tok": 204} {"draft_name": "draft-sharif-agent-transport-protocol", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"ATP is an asynchronous store-and-forward messaging protocol that enables autonomous AI agents to transmit themselves (including state, context, and capabilities) between agent runtimes across networks. The protocol uses SMTP-like semantics with cryptographic identity verification, trust scoring, and tamper-evident envelopes for secure agent mobility.\",\"n\":4,\"nn\":\"Notable originality in treating agents as self-contained mobile payloads with comprehensive state transfer, going beyond typical message-passing approaches\",\"m\":3,\"mn\":\"Well-defined protocol with clear operational model, security mechanisms, and interoperability claims, though lacks detailed message formats and implementation specifics\",\"o\":2,\"on\":\"Some conceptual overlap with mobile agent systems and secure messaging, but unique focus on AI agent mobility with modern crypto and trust mechanisms\",\"mo\":2,\"mon\":\"Single revision with future date suggests early stage; no evidence of working group adoption or broader community engagement yet\",\"r\":5,\"rn\":\"Directly addresses core AI agent communication challenges, particularly agent mobility and asynchronous operation in distributed environments\",\"c\":[\"A2A protocols\",\"Agent identity/auth\",\"Data formats/interop\"]}", "in_tok": 757, "out_tok": 251} {"draft_name": "draft-chuyi-nmrg-agentic-network-inference", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"Proposes an agentic network architecture to support AI agent interconnection and multi-level inference across cloud-edge-end deployments. Focuses on enabling low-latency, high-bandwidth communication for collaborative agent scenarios like smart homes, industrial robotics, and autonomous vehicles.\",\n \"n\": 3,\n \"nn\": \"Addresses the emerging need for agent-optimized network infrastructure, though the core networking concepts are evolutionary rather than revolutionary\",\n \"m\": 2,\n \"mn\": \"Presents problem statement and high-level architecture concepts but lacks detailed protocol specifications, mechanisms, or implementation guidance\",\n \"o\": 3,\n \"on\": \"Shares conceptual space with edge computing and distributed inference work, but the agent-centric networking focus provides some differentiation\",\n \"mo\": 1,\n \"mon\": \"Single draft with no apparent follow-up activity or working group engagement visible\",\n \"r\": 4,\n \"rn\": \"Directly addresses AI agent networking infrastructure needs, which is highly relevant to the emerging multi-agent ecosystem requirements\",\n \"c\": [\"A2A protocols\", \"Model serving/inference\", \"ML traffic mgmt\"]\n}", "in_tok": 742, "out_tok": 273} {"draft_name": "draft-zeng-mcp-network-measurement", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"Proposes using Model Context Protocol (MCP) to enable AI-assisted network measurement operations where network devices act as MCP servers and controllers as clients. The framework aims to provide natural language-driven network monitoring, fault diagnosis, and automated measurement workflows.\",\n \"n\": 3,\n \"nn\": \"Useful application of MCP to network measurement domain, though combining AI protocols with network management is becoming common\",\n \"m\": 2,\n \"mn\": \"Early-stage proposal with basic architecture description but lacks detailed protocol specifications and implementation details\",\n \"o\": 3,\n \"on\": \"Shares concepts with other AI-driven network management proposals and MCP application drafts\",\n \"mo\": 2,\n \"mon\": \"Single draft with no clear working group adoption or community engagement signals\",\n \"r\": 4,\n \"rn\": \"Directly relevant as it applies AI agent protocols to network operations with natural language interfaces\",\n \"c\": [\"Autonomous netops\", \"A2A protocols\", \"Human-agent interaction\"]\n}", "in_tok": 581, "out_tok": 245} {"draft_name": "draft-xu-agentic-overlay-network-architecture", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft proposes an overlay network architecture for autonomous software agents to discover and interact across administrative domains. It defines control-plane coordination through management nodes, registries, and discovery services, while keeping runtime execution decentralized between agent endpoints.\",\n \"n\": 4,\n \"nn\": \"Notable architectural approach separating control-plane coordination from runtime execution, with novel integration of trust anchors, semantic taxonomies, and enterprise service hubs in a unified overlay framework\",\n \"m\": 3,\n \"mn\": \"Defines clear architectural roles, functional boundaries, and operational workflows, but lacks detailed protocol specifications and concrete implementation examples\",\n \"o\": 2,\n \"on\": \"Some conceptual similarities to service mesh and overlay network approaches, but the agentic-specific combination of discovery, trust, and semantic coordination appears relatively unique\",\n \"mo\": 2,\n \"mn\": \"Single draft from 2026 with no visible working group adoption or community discussion yet\",\n \"r\": 5,\n \"rn\": \"Core AI agent networking topic addressing fundamental challenges of agent discovery, interoperability, and coordination across administrative boundaries\",\n \"c\": [\"A2A protocols\", \"Agent discovery/reg\", \"Agent identity/auth\", \"Policy/governance\"]\n}", "in_tok": 696, "out_tok": 290} {"draft_name": "draft-jovancevic-vdac", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"VDAC defines a cryptographic protocol for bilateral contracts between content publishers and automated agents regarding data access terms. The protocol enables verifiable agreement formation through signed offers, acceptances, and per-request binding references.\",\n \"n\": 4,\n \"nn\": \"Novel approach to formalizing publisher-agent data access agreements with cryptographic verification, addressing a gap in automated content access governance\",\n \"m\": 3,\n \"mn\": \"Defines core protocol mechanisms and contract formation process but appears to lack detailed implementation specifics and concrete examples\",\n \"o\": 2,\n \"on\": \"While related to web scraping and API access concepts, the cryptographic bilateral contract approach appears relatively unique in the IETF space\",\n \"mo\": 2,\n \"mon\": \"Future-dated draft (2026) suggests this is conceptual; no evidence of working group adoption or broader community engagement\",\n \"r\": 5,\n \"rn\": \"Directly addresses core AI agent operational needs for legitimate, verifiable data access relationships with content providers\",\n \"c\": [\"A2A protocols\", \"Agent identity/auth\", \"Policy/governance\"]\n}", "in_tok": 655, "out_tok": 269} {"draft_name": "draft-duda-dnsop-dns-did", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft proposes a DNS-based framework for privacy-preserving identity management that supports users, IoT devices, and AI agents through Self-Certifying Identifiers and trusted proxies. The approach uses DNSSEC-secured TXT records to bind public keys to identities while hiding real identities behind trusted intermediaries.\",\n \"n\": 3,\n \"nn\": \"Useful contribution applying existing DNS/DNSSEC infrastructure to identity management for AI agents, though the core concepts of self-certifying identifiers and trusted proxies are established\",\n \"m\": 2,\n \"mn\": \"Early sketch of framework with high-level concepts but lacks detailed protocol specifications, concrete examples, or implementation guidance\",\n \"o\": 4,\n \"on\": \"Significant overlap with existing DID methods, DNS-based identity systems, and privacy-preserving authentication schemes, though the specific combination for AI agents has some distinctiveness\",\n \"mo\": 2,\n \"mon\": \"Single revision draft with no indication of working group interest or broader community engagement\",\n \"r\": 4,\n \"rn\": \"Directly relevant to AI agents with explicit focus on automated, AI-driven interactions and identity management for AI agents alongside users and IoT devices\",\n \"c\": [\"Agent identity/auth\", \"A2A protocols\", \"Policy/governance\"]\n}", "in_tok": 651, "out_tok": 313} {"draft_name": "draft-emirdag-scitt-ai-agent-execution", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This IETF draft defines a SCITT profile for creating verifiable, tamper-evident records of AI agent actions through AgentInteractionRecords (AIRs). It maps SCITT roles to agent execution contexts and provides compliance mappings to major regulatory frameworks like EU AI Act and NIST AI RMF.\",\"n\":4,\"nn\":\"Notable originality in applying SCITT framework specifically to AI agent execution with comprehensive regulatory compliance mappings\",\"m\":4,\"mn\":\"Detailed specification with specific mechanisms like redaction receipts, hash chain integrity, and clear role mappings\",\"o\":2,\"on\":\"Minor similarities to general SCITT work but unique application to AI agent execution domain\",\"mo\":2,\"mon\":\"Single revision draft with no clear working group adoption yet\",\"r\":5,\"rn\":\"Core AI agent topic addressing critical transparency and accountability needs for autonomous systems\",\"c\":[\"Policy/governance\",\"Agent identity/auth\",\"Data formats/interop\"]}", "in_tok": 656, "out_tok": 212} {"draft_name": "draft-nennemann-wimse-ect", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft defines Execution Context Tokens (ECTs) as JWT-based extensions to WIMSE for tracking task execution in distributed agentic workflows. ECTs create signed records of individual tasks linked through DAGs and are transported via new HTTP headers alongside existing WIMSE identity mechanisms.\",\n \"n\": 3,\n \"nn\": \"Useful extension combining existing JWT/WIMSE patterns with workflow tracking - not groundbreaking but addresses real need for execution context in agent systems\",\n \"m\": 4,\n \"mn\": \"Well-detailed specification with clear protocol mechanics, HTTP header definitions, and appears implementation-ready though could benefit from more examples\",\n \"o\": 3,\n \"on\": \"Builds directly on WIMSE architecture and shares workflow concepts with some other distributed system drafts, but specific ECT approach is distinct\",\n \"mo\": 3,\n \"mon\": \"Recent draft in active development area, reasonable length suggests serious effort, but momentum unclear without revision history\",\n \"r\": 5,\n \"rn\": \"Directly addresses core AI agent infrastructure need for tracking distributed workflow execution with proper cryptographic attestation\",\n \"c\": [\"A2A protocols\", \"Agent identity/auth\", \"Data formats/interop\"]\n}", "in_tok": 557, "out_tok": 289} {"draft_name": "draft-hong-nmrg-agenticai-ps", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft presents the problem statement and motivation for introducing Agentic AI to network management, aiming for a paradigm shift where multiple autonomous AI agents collaborate to fully automate network operations. It identifies fundamental issues with existing technologies and proposes core areas requiring IETF/IRTF standardization.\",\"n\":2,\"nn\":\"Incremental application of existing agentic AI concepts to network management domain\",\"m\":1,\"mn\":\"Pure problem statement document with no technical mechanisms or protocols defined\",\"o\":3,\"on\":\"Shares general autonomous network management concepts with existing work but focuses specifically on agentic AI approach\",\"mo\":2,\"mon\":\"Single revision draft with no clear working group adoption or community engagement signals\",\"r\":5,\"rn\":\"Directly addresses core AI agent topics for network automation and multi-agent collaboration\",\"c\":[\"Autonomous netops\",\"A2A protocols\",\"Policy/governance\"]}", "in_tok": 569, "out_tok": 198} {"draft_name": "draft-sz-dmsc-iaip", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft specifies IAIP, a protocol for AI agent interconnection at Agent Gateways that enables dynamic discovery and routing based on semantic intent rather than static addressing. The protocol defines mechanisms for agent registration, capability advertisement, intent resolution and matching to facilitate agent-to-agent communication.\",\"n\":4,\"nn\":\"Intent-based agent interconnection represents a notable departure from traditional network addressing paradigms, introducing semantic matching and capability-based routing for AI agents\",\"m\":3,\"mn\":\"Defines core protocol mechanisms including registration, validation, and routing but lacks detailed message formats, state machines, or implementation specifics\",\"o\":2,\"on\":\"While touching on service discovery concepts, the intent-based semantic routing approach for AI agents appears relatively unique in IETF space\",\"mo\":2,\"mon\":\"Single draft submission with no clear working group adoption or community discussion visible\",\"r\":5,\"rn\":\"Directly addresses core AI agent networking challenges including discovery, capability matching, and intent-based communication - highly relevant to AI agent infrastructure\",\"c\":[\"A2A protocols\",\"Agent discovery/reg\"]}", "in_tok": 596, "out_tok": 235} {"draft_name": "draft-ramadan-mboned-sonar", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"SONAR proposes a protocol for verifiable multicast delivery claims using cryptoeconomic accountability and statistical sampling to detect cheating without trusted intermediaries. It combines ALTA-based packet authentication with VRF sampling and blockchain attestations to verify content delivery at scale with minimal overhead.\",\"n\":4,\"nn\":\"Novel combination of multicast verification, cryptoeconomic incentives, and statistical sampling using German Tank Problem - creative approach to a challenging distributed systems problem\",\"m\":3,\"mn\":\"Well-defined protocol with clear mechanisms and mathematical foundations, but lacks detailed packet formats, state machines, and implementation specifics\",\"o\":2,\"on\":\"Unique approach combining multicast networking with cryptoeconomic verification - some overlap with general blockchain attestation concepts but novel application domain\",\"mo\":2,\"mon\":\"Single draft submission with no apparent working group adoption or community discussion yet\",\"r\":1,\"rn\":\"This is a networking protocol for multicast delivery verification, not related to AI agents or artificial intelligence applications\",\"c\":[\"Other AI/agent\"]}", "in_tok": 716, "out_tok": 224} {"draft_name": "draft-li-individual-inip", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"INIP is a protocol for performing lightweight ML inference directly in network data plane devices like switches and SmartNICs using a two-tier architecture with centralized model management and distributed execution. The protocol uses IPv4-based packet formats and match-action tables to enable high-speed inference without impacting core forwarding functions.\",\n \"n\": 4,\n \"nn\": \"Novel approach to push ML inference into network data plane devices using match-action tables, representing a creative intersection of networking hardware capabilities and AI inference\",\n \"m\": 4,\n \"mn\": \"Detailed 17-page specification covering packet formats, architecture, control plane responsibilities, and CDN-like scheduling mechanisms with sufficient detail for implementation\",\n \"o\": 2,\n \"on\": \"While in-network computing exists, the specific focus on ML inference using match-action tables in data plane devices appears to have limited direct overlap with existing IETF work\",\n \"mo\": 2,\n \"mn\": \"Individual draft with recent timestamp but unclear adoption status or working group interest in this specific approach to in-network inference\",\n \"r\": 5,\n \"rn\": \"Directly addresses AI model serving and inference with a novel network-centric approach that could enable new classes of AI-powered network applications\",\n \"c\": [\"Model serving/inference\", \"Autonomous netops\"]\n}", "in_tok": 694, "out_tok": 314} {"draft_name": "draft-zl-agents-networking-scenarios", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft outlines networking scenarios for AI agents in enterprise and home broadband environments, focusing on how these differ from 6G and general Internet use cases. The document appears to be primarily scenario-based rather than proposing specific technical solutions.\",\"n\":2,\"nn\":\"Incremental work applying existing agent concepts to specific network environments\",\"m\":2,\"mn\":\"Early stage draft with scenario descriptions but lacks detailed technical mechanisms\",\"o\":3,\"on\":\"Shares general agent networking concepts with other drafts but focuses on specific deployment environments\",\"mo\":2,\"mon\":\"Single revision with no clear working group activity or community engagement\",\"r\":4,\"rn\":\"Directly addresses AI agent networking in practical deployment scenarios\",\"c\":[\"A2A protocols\",\"Autonomous netops\",\"Agent discovery/reg\"]}", "in_tok": 515, "out_tok": 171} {"draft_name": "draft-vandemeent-tibet-provenance", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"TIBET defines a cryptographically-linked provenance chain protocol for tracking interactions between autonomous agents, humans, and automated processes using four-dimensional tokens. The protocol provides hash-chained, signed evidence trails with JSON serialization and integration points for companion agent protocols.\",\n \"n\": 4,\n \"nn\": \"Notable originality in four-dimensional provenance model (content, references, context, intent) and comprehensive cryptographic chaining for multi-agent interactions\",\n \"m\": 4,\n \"mn\": \"Detailed specification with token data model, chain semantics, canonicalization rules, verification procedures, and defined integration points\",\n \"o\": 2,\n \"on\": \"Minor similarities to general blockchain/audit trail concepts but unique multi-dimensional provenance approach for agent interactions\",\n \"mo\": 3,\n \"mon\": \"Active development indicated by integration with multiple companion protocols (JIS, UPIP, RVP, AINS) suggesting coordinated effort\",\n \"r\": 5,\n \"rn\": \"Core AI agent topic addressing critical provenance and accountability challenges in autonomous agent interactions\",\n \"c\": [\"A2A protocols\", \"AI safety/alignment\", \"Data formats/interop\"]\n}", "in_tok": 658, "out_tok": 284} {"draft_name": "draft-mw-spice-intent-chain", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft defines a cryptographically verifiable intent chain to track content provenance through AI agent workflows, complementing the actor chain for delegation provenance. It uses Merkle roots in OAuth tokens to efficiently store tamper-evident logs of content transformations across processing stages.\",\n \"n\": 4,\n \"nn\": \"Notable originality in combining content provenance tracking with cryptographic verification specifically for AI agent workflows, extending beyond traditional delegation models\",\n \"m\": 3,\n \"mn\": \"Defines the protocol mechanism with clear abstract and relationship to companion draft, though implementation details and examples not fully evident from abstract\",\n \"o\": 2,\n \"on\": \"While building on the companion actor-chain draft, the content provenance focus appears to have minor similarities to existing work rather than significant overlap\",\n \"mo\": 3,\n \"mon\": \"Active development evidenced by 2026 date and companion draft relationship, though working group adoption status unclear\",\n \"r\": 5,\n \"rn\": \"Core AI agent topic directly addressing content provenance, tampering, and governance challenges fundamental to autonomous AI systems\",\n \"c\": [\"Agent identity/auth\", \"Policy/governance\", \"AI safety/alignment\"]\n}", "in_tok": 662, "out_tok": 283} {"draft_name": "draft-ietf-tls-deprecate-obsolete-kex", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This IETF draft deprecates obsolete key exchange methods (finite field DH and RSA) in TLS/DTLS 1.2, updating multiple RFCs to change recommendations to MUST NOT or SHOULD NOT use these cipher suites. The document focuses solely on TLS 1.2 since earlier versions are already deprecated and TLS 1.3 doesn't use these problematic algorithms.\",\"n\":1,\"nn\":\"This is a straightforward deprecation of well-known obsolete cryptographic methods - no novel technical contribution, just necessary housekeeping\",\"m\":4,\"mn\":\"Detailed specification with clear normative language updates to existing RFCs, ready for implementation with specific cipher suite deprecations listed\",\"o\":3,\"on\":\"Shares concepts with other TLS security update drafts but addresses a specific set of key exchange methods not covered elsewhere\",\"mo\":4,\"mon\":\"Strong IETF working group momentum as this addresses widely recognized security issues in deployed TLS implementations\",\"r\":1,\"rn\":\"This is purely about TLS cryptographic protocol updates with no connection to AI, agents, or machine learning - complete false positive\",\"c\":[\"Other AI/agent\"]}", "in_tok": 789, "out_tok": 262} {"draft_name": "draft-abaris-aicdh", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft proposes an HTTP response header 'AI-Disclosure' to indicate when web content was generated or assisted by AI. The header uses structured field syntax to provide machine-readable metadata about AI involvement for user agents, bots, and archiving systems.\",\"n\":3,\"nn\":\"Useful contribution addressing a timely need for AI content transparency, though the concept of disclosure headers is relatively straightforward\",\"m\":4,\"mn\":\"Detailed specification with structured field syntax, examples, and clear implementation guidance for HTTP headers\",\"o\":2,\"on\":\"Minor similarities to general content metadata approaches but unique focus on AI disclosure makes it largely distinct\",\"mo\":2,\"mon\":\"Single revision draft without clear working group adoption or broader community engagement\",\"r\":4,\"rn\":\"Directly relevant to AI transparency and human-agent interaction, addressing important disclosure needs for AI-generated content\",\"c\":[\"Human-agent interaction\",\"Policy/governance\",\"Data formats/interop\"]}", "in_tok": 543, "out_tok": 206} {"draft_name": "draft-eckert-anima-acp-free-ani", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft proposes a lightweight alternative to ANIMA's Autonomic Networking Infrastructure by eliminating the complex ACP component while retaining certificates, GRASP, and BRSKI protocols. The approach uses application-level Autonomic Software Agents with proxying capabilities to achieve most ANI benefits without requiring changes to router/switch control planes.\",\n \"n\": 3,\n \"nn\": \"Useful simplification of existing ANIMA framework, trading some security guarantees for implementation practicality\",\n \"m\": 3,\n \"mn\": \"Defines the architecture and protocol interactions but lacks detailed implementation specifications and examples\",\n \"o\": 4,\n \"on\": \"Significant overlap with existing ANIMA drafts, essentially proposing a subset/variation of established ANI concepts\",\n \"mo\": 2,\n \"mon\": \"Single revision with future date suggests early stage work, unclear community engagement within ANIMA working group\",\n \"r\": 4,\n \"rn\": \"Directly relevant to autonomous network operations and distributed automation agents, core networking AI topic\",\n \"c\": [\"Autonomous netops\", \"Agent discovery/reg\", \"A2A protocols\"]\n}", "in_tok": 699, "out_tok": 277} {"draft_name": "draft-jennings-mcp-over-moqt", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft specifies how to transport Model Context Protocol (MCP) messages over Media over QUIC Transport (MOQT), enabling efficient publish-subscribe delivery for AI language model integrations with external data sources. It defines the mapping between MCP messages and MOQT objects along with session establishment procedures.\",\"n\":2,\"nn\":\"Incremental work adapting existing MCP to run over MOQT transport - straightforward protocol layering\",\"m\":3,\"mn\":\"Defines complete protocol mapping and procedures but lacks implementation details and examples\",\"o\":3,\"on\":\"Shares transport concepts with other QUIC-based drafts and relates to MCP specifications\",\"mo\":2,\"mon\":\"Single revision with 2026 date suggests early stage without demonstrated community engagement\",\"r\":4,\"rn\":\"Directly relevant to AI agents as MCP is specifically designed for language model application integration\",\"c\":[\"A2A protocols\",\"Data formats/interop\"]}", "in_tok": 576, "out_tok": 210} {"draft_name": "draft-messous-eat-ai", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft defines an EAT profile for remotely attesting autonomous AI agents, specifying claims for model integrity, training data provenance, and inference policies. It includes 5G/6G extensions and uses CWT/JWT encoding within the IETF RATS architecture.\",\n \"n\": 4,\n \"nn\": \"Notable originality in applying remote attestation specifically to AI agents with comprehensive claims covering model parameters, training provenance, and inference constraints\",\n \"m\": 2,\n \"mn\": \"Early sketch with basic profile definition but lacks detailed implementation specifics, test vectors, or worked examples\",\n \"o\": 2,\n \"on\": \"Minor similarities to general EAT work but unique AI-specific application with novel claims structure\",\n \"mo\": 2,\n \"mon\": \"Single revision with no evidence of working group adoption or broader community engagement\",\n \"r\": 5,\n \"rn\": \"Core AI agent topic addressing critical trust and attestation challenges for autonomous AI systems across domains\",\n \"c\": [\"Agent identity/auth\", \"AI safety/alignment\", \"Data formats/interop\"]\n}", "in_tok": 618, "out_tok": 264} {"draft_name": "draft-ietf-rats-coserv", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"CoSERV defines a CBOR-based query language for discovering and retrieving endorsements and reference values in remote attestation systems. The draft specifies structured formats using CDDL to enable efficient interoperability between verifiers and endorsement providers.\",\"n\":2,\"nn\":\"incremental extension of existing RATS architecture with standard query mechanisms\",\"m\":4,\"mn\":\"detailed 69-page specification with CDDL definitions and CBOR serialization ready for implementation\",\"o\":3,\"on\":\"builds directly on RATS architecture and shares attestation concepts with related security drafts\",\"mo\":4,\"mon\":\"IETF working group draft with substantial page count indicating active development\",\"r\":2,\"rn\":\"focuses on attestation infrastructure rather than AI agents, though could support agent verification scenarios\",\"c\":[\"Agent identity/auth\",\"Data formats/interop\"]}", "in_tok": 585, "out_tok": 191} {"draft_name": "draft-zhao-anima-automatic-congestion-relief", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft proposes an automatic congestion relief mechanism using intelligent traffic analysis and dynamic regulation for network self-healing. It focuses on responding to congestion from fiber optic failures through real-time intelligent adaptation.\",\"n\":2,\"nn\":\"Incremental application of existing AI traffic management concepts to congestion relief\",\"m\":2,\"mn\":\"Early sketch with abstract concepts but lacks detailed protocol specification\",\"o\":4,\"on\":\"Significant overlap with existing traffic engineering and AI-based network management approaches\",\"mo\":2,\"mon\":\"Single revision with no evidence of working group interest or community engagement\",\"r\":3,\"rn\":\"Partially relevant as it applies AI for autonomous network operations but focuses narrowly on congestion\",\"c\":[\"ML traffic mgmt\",\"Autonomous netops\"]}", "in_tok": 507, "out_tok": 167} {"draft_name": "draft-jimenez-t2trg-iot-agent", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft defines an architecture for LLM-based agents to autonomously discover and interact with IoT devices using CoAP and hypermedia patterns. The approach enables agents to decompose high-level user intents into device operations without pre-configured knowledge, relying on in-band discovery and semantic metadata.\",\n \"n\": 4,\n \"nn\": \"Notable originality in applying LLM agents to constrained IoT environments with hypermedia-driven autonomous discovery - bridges AI agents with resource-constrained networking in a novel way\",\n \"m\": 3,\n \"mn\": \"Defines clear protocol mechanisms and architecture with CoAP/CoRE integration, though implementation details and concrete examples could be more extensive for full specification readiness\",\n \"o\": 2,\n \"on\": \"Minor similarities to general IoT discovery work, but the combination of LLM agents with constrained RESTful environments appears largely unique\",\n \"mo\": 2,\n \"mon\": \"Single revision draft with future 2026 date suggests early stage development without demonstrated working group traction yet\",\n \"r\": 5,\n \"rn\": \"Core AI agent topic directly addressing autonomous operation in networked environments - exemplifies intelligent agents making independent decisions in constrained systems\",\n \"c\": [\"Autonomous netops\", \"Agent discovery/reg\", \"Data formats/interop\"]\n}", "in_tok": 621, "out_tok": 318} {"draft_name": "draft-ffm-rats-cca-token", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft specifies the attestation token format for Arm's Confidential Compute Architecture (CCA), defining a profile of Entity Attestation Token (EAT) for trustworthiness assessment. It describes the claims structure, serialization, and cryptographic protection for CCA-compliant systems.\",\"n\":2,\"nn\":\"Incremental work creating a hardware-specific profile of existing EAT standard\",\"m\":4,\"mn\":\"Detailed specification with clear structure and semantics for implementation\",\"o\":3,\"on\":\"Shares concepts with EAT and other attestation token specifications\",\"mo\":3,\"mon\":\"Active development as informational draft for Arm architecture interoperability\",\"r\":2,\"rn\":\"Tangentially related to AI/agents through confidential computing security foundations\",\"c\":[\"Agent identity/auth\"]}", "in_tok": 671, "out_tok": 184} {"draft_name": "draft-stone-vcap", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"VCAP specifies a protocol for autonomous AI agents to conduct verified financial transactions using cryptographic proof of work delivery and escrow mechanisms. It defines message formats, state machines, and verification contracts to enable trustless commerce between agents with automatic payment settlement.\",\n \"n\": 4,\n \"nn\": \"Novel approach to AI agent commerce with cryptographic verification - no existing standards address autonomous agent financial transactions with proof-of-work delivery\",\n \"m\": 4,\n \"mn\": \"Detailed specification with defined message formats, state machines, cryptographic bindings, and callback contracts - appears implementation-ready\",\n \"o\": 1,\n \"on\": \"Unique focus on verified commerce for AI agents - no significant overlap with existing IETF work on agent protocols or commerce standards\",\n \"mo\": 2,\n \"mon\": \"Single draft submission with no clear working group adoption or community feedback yet\",\n \"r\": 5,\n \"rn\": \"Directly addresses core AI agent infrastructure need for trustless financial transactions and automated commerce verification\",\n \"c\": [\"A2A protocols\", \"Agent identity/auth\", \"Data formats/interop\"]\n}", "in_tok": 623, "out_tok": 269} {"draft_name": "draft-ounsworth-lamps-x509-ar", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft defines X.509 certificate extensions to embed attestation results, primarily for CSR attestation where claims about device/system trustworthiness are included in certificates. It enables CAs to appraise evidence and embed attestation results for relying parties to evaluate trustworthiness consistently.\",\"n\":3,\"nn\":\"Useful contribution applying established attestation concepts to X.509 certificates, though the core idea of embedding attestation data in certificates is somewhat expected\",\"m\":4,\"mn\":\"Well-defined specification with detailed extension structures, ASN.1 definitions, and clear processing requirements, though could benefit from more implementation examples\",\"o\":3,\"on\":\"Shares concepts with existing attestation and PKI work, particularly RATS architecture and X.509 extension mechanisms, but creates a specific bridge between these domains\",\"mo\":2,\"mn\":\"Recent draft with limited revision history, unclear working group adoption status or broader community engagement\",\"r\":2,\"rn\":\"Tangentially related to AI/agents as attestation could apply to AI systems, but primarily focuses on general device/system attestation rather than AI-specific concerns\",\"c\":[\"Agent identity/auth\"]}", "in_tok": 609, "out_tok": 252} {"draft_name": "draft-li-semantic-routing-architecture", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft proposes a Semantic Routing Architecture that uses application-layer semantics like intent vectors and trust scores to intelligently route communications between AI agents, moving beyond traditional IP routing. The architecture aims to transform networks from passive transport to active collaboration substrates for multi-agent coordination.\",\n \"n\": 4,\n \"nn\": \"Notable originality in applying semantic-driven routing specifically to AI agent communication, moving beyond traditional network routing paradigms\",\n \"m\": 2,\n \"mn\": \"Early sketch level - introduces concepts but lacks detailed protocol specifications, implementation details, or concrete examples\",\n \"o\": 2,\n \"on\": \"Minor similarities to service mesh and intent-based networking concepts, but unique application to AI agent semantics\",\n \"mo\": 2,\n \"mon\": \"Single revision with future date suggests early stage development without demonstrated community engagement\",\n \"r\": 5,\n \"rn\": \"Core AI agent topic addressing fundamental infrastructure needs for multi-agent communication and coordination\",\n \"c\": [\"A2A protocols\", \"Agent discovery/reg\", \"Autonomous netops\"]\n}", "in_tok": 580, "out_tok": 257} {"draft_name": "draft-bradleylundberg-cfrg-arkg", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"ARKG enables delegation of asymmetric public key generation without exposing private keys, supporting applications like pseudonymous keys for user agents and ephemeral keys for enhanced forward secrecy. The draft provides a generic algorithm specification, instantiation formulae, and concrete implementations.\",\n \"n\": 3,\n \"nn\": \"Useful cryptographic contribution that addresses a real need in distributed systems, though builds on established asymmetric cryptography concepts\",\n \"m\": 4,\n \"mn\": \"Detailed specification with abstract algorithm, concrete instantiation formulae, and fully specified instances ready for implementation\",\n \"o\": 2,\n \"on\": \"While related to general key management work, the specific ARKG approach and application to agent scenarios appears relatively unique\",\n \"mo\": 2,\n \"mon\": \"Appears to be early stage work without clear working group adoption or broader community engagement yet\",\n \"r\": 4,\n \"rn\": \"Directly relevant to AI agents needing pseudonymous identities, secure messaging, and paired authentication devices without shared private keys\",\n \"c\": [\"Agent identity/auth\", \"A2A protocols\"]\n}", "in_tok": 644, "out_tok": 270} {"draft_name": "draft-ar-emu-pqc-eapaka", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft proposes enhancing EAP-AKA' Forward Secrecy with hybrid post-quantum cryptography to protect against future quantum computer threats. It aims to replace the traditional ECDHE key exchange with quantum-resistant algorithms while maintaining backward compatibility.\",\n \"n\": 2,\n \"nn\": \"Incremental application of established PQ/T hybrid approach to existing EAP-AKA' protocol - straightforward quantum-resistant upgrade\",\n \"m\": 2,\n \"mn\": \"Early sketch stage - provides motivation and general approach but lacks detailed protocol specification or implementation guidance\",\n \"o\": 4,\n \"on\": \"Significant conceptual overlap with other post-quantum authentication protocols and hybrid cryptography drafts in IETF\",\n \"mo\": 3,\n \"mon\": \"Single draft version addressing timely quantum-readiness concerns in mobile authentication, moderate development activity expected\",\n \"r\": 1,\n \"rn\": \"Pure cryptographic protocol enhancement for mobile network authentication - no AI or agent-related functionality\",\n \"c\": [\"Other AI/agent\"]\n}", "in_tok": 720, "out_tok": 261} {"draft_name": "draft-happel-structured-email-trust", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft addresses trust and security considerations for structured email, providing recommendations for message user agents on handling structured data in email messages. It focuses on security implications when emails contain machine-readable structured content that could be processed by automated systems.\",\"n\":2,\"nn\":\"Incremental work addressing known security concerns around structured data in email\",\"m\":2,\"mn\":\"Early stage draft with basic problem outline but limited technical detail\",\"o\":3,\"on\":\"Shares concepts with email security and structured data standards\",\"mo\":2,\"mon\":\"Single draft with no clear working group adoption\",\"r\":3,\"rn\":\"Partially relevant as structured email could enable agent communication but not explicitly focused on AI agents\",\"c\":[\"A2A protocols\",\"Data formats/interop\",\"Policy/governance\"]}", "in_tok": 476, "out_tok": 167} {"draft_name": "draft-ietf-sshm-mlkem-hybrid-kex", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This IETF draft defines hybrid key exchange methods combining quantum-resistant ML-KEM with traditional ECDH for SSH transport protocol. It addresses post-quantum cryptography readiness by enabling dual protection against both classical and quantum attacks.\",\"n\":2,\"nn\":\"Incremental work applying established PQ/T hybrid approach to SSH, following patterns from TLS implementations\",\"m\":4,\"mn\":\"Detailed specification with concrete protocol definitions, algorithm identifiers, and wire format specifications\",\"o\":3,\"on\":\"Shares hybrid PQ/T concepts with TLS drafts and other SSH cryptographic extensions but SSH-specific implementation\",\"mo\":4,\"mon\":\"Strong working group adoption with clear timeline to 2026, addressing urgent post-quantum transition needs\",\"r\":1,\"rn\":\"Pure cryptographic protocol specification with no AI or agent-related functionality\",\"c\":[\"Other AI/agent\"]}", "in_tok": 552, "out_tok": 195} {"draft_name": "draft-ietf-core-oscore-edhoc", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft specifies how to use the EDHOC authenticated key exchange protocol over CoAP to establish OSCORE security contexts for constrained environments. It includes optimizations to combine EDHOC execution with the first OSCORE transaction to reduce round trips.\",\"n\":2,\"nn\":\"Incremental work combining existing protocols (EDHOC, CoAP, OSCORE) with some optimization contributions\",\"m\":4,\"mn\":\"Detailed specification with concrete mechanisms, message flows, and implementation guidance\",\"o\":3,\"on\":\"Builds on established CoAP/OSCORE work but addresses specific integration aspects\",\"mo\":4,\"mon\":\"IETF working group draft suggesting active development and community interest\",\"r\":1,\"rn\":\"This is about constrained device security protocols, not AI/agents - appears to be a false positive\",\"c\":[\"Other AI/agent\"]}", "in_tok": 663, "out_tok": 195} {"draft_name": "draft-it-aipref-attachment", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft defines a mechanism for content creators to signal their preferences about how automated systems should consume their content via HTTP headers. It updates RFC 9309 to include usage preferences as part of content acquisition protocols.\",\"n\":2,\"nn\":\"Incremental extension of existing HTTP preference mechanisms to address AI content consumption - useful but builds on established patterns\",\"m\":3,\"mn\":\"Defines a protocol mechanism with clear structure, though would benefit from more detailed examples and implementation guidance\",\"o\":3,\"on\":\"Shares concepts with robots.txt, content licensing mechanisms, and other HTTP preference frameworks\",\"mo\":2,\"mon\":\"Single revision draft with limited visible community engagement or working group adoption\",\"r\":4,\"rn\":\"Directly addresses the important issue of AI systems respecting content creator preferences during automated content consumption\",\"c\":[\"Policy/governance\",\"Data formats/interop\"]}", "in_tok": 513, "out_tok": 186} {"draft_name": "draft-traffic-analysis-and-network-mode-mapping", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft proposes using large language models for network traffic classification and resource allocation mapping. The framework includes pre-training on packet capture data to generate traffic representations and dynamic mapping for resource allocation strategies.\",\n \"n\": 3,\n \"nn\": \"Useful contribution applying LLMs to traffic analysis, though LLM-based network management is an established research direction\",\n \"m\": 2,\n \"mn\": \"Early sketch level - lacks detailed protocol specifications, implementation details, or concrete examples of the proposed framework\",\n \"o\": 4,\n \"on\": \"Significant overlap with existing ML-based traffic classification and QoS management approaches, though LLM application adds some differentiation\",\n \"mo\": 1,\n \"mon\": \"Future-dated draft (2026) with no evidence of current development activity or community engagement\",\n \"r\": 4,\n \"rn\": \"Directly relevant as it applies AI/ML techniques to network operations, representing autonomous network management capabilities\",\n \"c\": [\"ML traffic mgmt\", \"Autonomous netops\"]\n}", "in_tok": 605, "out_tok": 250} {"draft_name": "draft-liu-agent-context-protocol", "model": "claude-sonnet-4-20250514", "raw": "```json\n{\n \"s\": \"Defines a standardized protocol for AI agents to exchange context information with each other. Proposes message formats and communication mechanisms for inter-agent context sharing.\",\n \"n\": 3,\n \"nn\": \"Addresses a recognized need for agent interoperability but builds on established messaging concepts\",\n \"m\": 3,\n \"mn\": \"8 pages suggests defined protocol structure but may lack comprehensive implementation details\",\n \"o\": 3,\n \"on\": \"Likely shares messaging and protocol concepts with other agent communication proposals\",\n \"mo\": 2,\n \"mon\": \"Recent draft date but no indication of working group adoption or broader community engagement\",\n \"r\": 5,\n \"rn\": \"Directly addresses core challenge of AI agent communication and context sharing\",\n \"c\": [\"A2A protocols\", \"Data formats/interop\"]\n}\n```", "in_tok": 458, "out_tok": 210} {"draft_name": "draft-madhavan-aipref-displaybasedpref", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"Proposes a vocabulary for content creators to declare preferences about how search engines and AI crawlers can use their collected content. The draft defines structured declarations to express restrictions or permissions for content usage by automated systems.\",\"n\":2,\"nn\":\"Incremental extension of existing robots.txt concepts to AI/ML context\",\"m\":2,\"mn\":\"Early sketch with basic vocabulary concepts but lacks detailed specification\",\"o\":3,\"on\":\"Shares concepts with existing web crawling standards and emerging AI content usage frameworks\",\"mo\":2,\"mon\":\"Single revision with uncertain adoption trajectory\",\"r\":4,\"rn\":\"Directly addresses current concerns about AI training data collection and usage rights\",\"c\":[\"Policy/governance\",\"Data formats/interop\"]}", "in_tok": 513, "out_tok": 158} {"draft_name": "draft-yu-dmsc-ai-agent-use-cases-in-6g", "model": "claude-sonnet-4-20250514", "raw": "```json\n{\n \"s\": \"This draft outlines use cases for AI agents in 6G networks, building on 3GPP technical reports and focusing on operator requirements. It appears to be a requirements and use case document rather than a technical specification.\",\n \"n\": 2,\n \"nn\": \"Incremental work that largely references existing 3GPP studies on 6G use cases, with limited novel contributions beyond organizing AI agent scenarios for network operators\",\n \"m\": 2,\n \"mn\": \"Early stage document that presents use cases and requirements but lacks detailed technical mechanisms, protocols, or implementation guidance\",\n \"o\": 4,\n \"on\": \"Significant overlap with 3GPP TR 22.870 and other 6G network AI integration work, as explicitly acknowledged in the abstract\",\n \"mo\": 1,\n \"mon\": \"Single draft submission with no evidence of working group adoption or community engagement in IETF context\",\n \"r\": 3,\n \"rn\": \"Partially relevant as it discusses AI agents in network context, but focuses more on telecommunications use cases than general AI agent protocols or interoperability\",\n \"c\": [\"Autonomous netops\", \"Policy/governance\", \"Other AI/agent\"]\n}\n```", "in_tok": 538, "out_tok": 289} {"draft_name": "draft-liu-agent-operation-authorization", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft proposes a JWT-based framework for authorizing AI agent operations through structured request and confirmation phases. The system aims to prevent unauthorized agent actions by requiring cryptographic verification of user intent before agents can execute operations.\",\"n\":3,\"nn\":\"Useful contribution applying existing JWT mechanisms to the emerging problem of AI agent authorization, though the core concept of structured authorization tokens is well-established\",\"m\":4,\"mn\":\"Detailed specification with clear two-phase protocol definition, JWT structure, and operational procedures, appears implementation-ready\",\"o\":2,\"on\":\"Minor conceptual similarities to general authorization frameworks but specific application to AI agent operations is relatively unique\",\"mo\":2,\"mon\":\"Single revision with 2026 date suggests early stage development, no evident working group adoption yet\",\"r\":5,\"rn\":\"Directly addresses core AI agent security concerns around unauthorized operations and verifiable delegation of authority\",\"c\":[\"Agent identity/auth\",\"Human-agent interaction\",\"AI safety/alignment\"]}", "in_tok": 599, "out_tok": 210} {"draft_name": "draft-ramakrishna-satp-views-addresses", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft defines a protocol for DLT systems to project views of virtual assets to external parties with access control and verification capabilities. It enables DLT-neutral addressing and retrieval of asset states, locks, and computed functions across blockchain boundaries.\",\n \"n\": 3,\n \"nn\": \"Useful contribution addressing asset interoperability across DLT systems, though builds on established blockchain interop concepts\",\n \"m\": 3,\n \"mn\": \"Defines protocol mechanisms and view concepts but appears to lack detailed implementation specifics and examples\",\n \"o\": 3,\n \"on\": \"Shares concepts with blockchain interoperability and cross-chain asset transfer protocols in the space\",\n \"mo\": 2,\n \"mon\": \"Single revision with future date suggests early stage development without established community engagement\",\n \"r\": 2,\n \"rn\": \"Focuses on blockchain asset transfer rather than AI agents, though external parties could theoretically include automated agents\",\n \"c\": [\"Other AI/agent\", \"Data formats/interop\"]\n}", "in_tok": 751, "out_tok": 246} {"draft_name": "draft-ainp-protocol", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"Proposes AINP, a semantic communication protocol for AI agents that replaces traditional network routing with capability-based discovery and intent exchange. The protocol enables autonomous agent negotiation and structured communication with cryptographic security.\",\n \"n\": 4,\n \"nn\": \"Notable originality in replacing fundamental networking primitives (location-based routing, byte streams) with semantic equivalents for AI agent communication\",\n \"m\": 3,\n \"mn\": \"Defines core protocol mechanisms and semantic routing concepts but likely needs more detailed specification and examples for implementation\",\n \"o\": 2,\n \"on\": \"While agent communication exists, the semantic routing and intent-based approach appears relatively unique in the networking protocol space\",\n \"mo\": 2,\n \"mon\": \"Single revision of a new draft without clear working group adoption or community engagement signals\",\n \"r\": 5,\n \"rn\": \"Directly addresses core AI agent networking needs with semantic communication, capability discovery, and autonomous negotiation\",\n \"c\": [\"A2A protocols\", \"Agent discovery/reg\", \"Agent identity/auth\"]\n}", "in_tok": 559, "out_tok": 255} {"draft_name": "draft-jiang-intent-security", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft analyzes security threats in intent-based agentic systems where high-level user goals are translated into executable actions across multiple entities. It provides a reference model for understanding attack surfaces like privilege escalation and intent drift in multi-hop agent processing chains.\",\"n\":3,\"nn\":\"Useful contribution addressing emerging security concerns in agent systems, though intent-based computing concepts exist in other domains\",\"m\":2,\"mn\":\"Early sketch providing threat analysis and requirements but lacks concrete protocols or detailed security mechanisms\",\"o\":2,\"on\":\"Some conceptual overlap with general agent security work but specifically focused on intent-based request processing chains\",\"mo\":2,\"mon\":\"Single revision draft with no clear working group engagement or community discussion visible\",\"r\":5,\"rn\":\"Directly addresses core security challenges in AI agent systems, highly relevant to emerging agentic architectures\",\"c\":[\"AI safety/alignment\",\"Agent identity/auth\",\"Policy/governance\"]}", "in_tok": 642, "out_tok": 202} {"draft_name": "draft-liu-ztcpp-zt-problem-statement", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft presents a problem statement for standardizing zero trust interoperability protocols in IETF, arguing that existing high-level architectural guidance lacks the open, interoperable frameworks needed for multi-vendor zero trust environments. It outlines use cases and discusses potential scope for IETF standardization work versus other standards bodies.\",\"n\":2,\"nn\":\"Incremental work identifying known gaps in zero trust standardization without proposing novel technical approaches\",\"m\":1,\"mn\":\"Pure problem statement document with no technical solutions, protocols, or mechanisms defined\",\"o\":4,\"on\":\"Significant conceptual overlap with existing zero trust architecture work and various IETF security protocols, though focused on interoperability gaps\",\"mo\":2,\"mon\":\"Single revision draft with numerous typos suggesting limited review and development activity\",\"r\":2,\"rn\":\"Only tangentially related to AI/agents through brief mention of 'intelligent agents' as one use case, primarily focused on general zero trust security\",\"c\":[\"Policy/governance\"]}", "in_tok": 738, "out_tok": 220} {"draft_name": "draft-nederveld-adl", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"ADL proposes a JSON-based standard for describing AI agents including their capabilities, tools, permissions, and security requirements. The draft defines the structure and semantics for agent description documents to enable discovery, interoperability, and lifecycle management across platforms.\",\"n\":2,\"nn\":\"Incremental approach applying standard metadata description patterns to AI agents - useful but not particularly novel\",\"m\":4,\"mn\":\"Detailed specification with clear structure definitions, semantics, and media type registration - appears implementation-ready\",\"o\":3,\"on\":\"Shares concepts with other agent description and service discovery formats, but specific to AI agents\",\"mo\":2,\"mon\":\"Single draft with no clear working group adoption or community feedback visible\",\"r\":5,\"rn\":\"Directly addresses core AI agent interoperability and standardization needs\",\"c\":[\"Data formats/interop\",\"Agent discovery/reg\",\"Agent identity/auth\"]}", "in_tok": 578, "out_tok": 194} {"draft_name": "draft-singh-psi", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"Proposes a cryptographic protocol for AI regulatory compliance verification using zero-knowledge proofs and multi-party computation. Organizations can prove compliance with regulations like the EU AI Act without revealing proprietary model details.\",\n \"n\": 4,\n \"nn\": \"Novel application of established cryptographic primitives to AI regulatory compliance, addressing a genuine gap in privacy-preserving compliance verification\",\n \"m\": 3,\n \"mn\": \"Defines specific protocol components and cryptographic mechanisms but lacks detailed implementation specifics and test vectors for a 5-page draft\",\n \"o\": 2,\n \"on\": \"While using established cryptographic techniques, the specific application to AI compliance verification appears relatively unique\",\n \"mo\": 2,\n \"mon\": \"Single revision with no clear indication of working group adoption or broader community engagement\",\n \"r\": 4,\n \"rn\": \"Directly addresses AI governance and compliance verification, highly relevant to AI systems regulation and deployment\",\n \"c\": [\"Policy/governance\", \"AI safety/alignment\", \"Agent identity/auth\"]\n}", "in_tok": 639, "out_tok": 249} {"draft_name": "draft-gaikwad-south-authorization", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"SOUTH defines a stochastic authorization protocol that allows servers to make probabilistic access control decisions based on uncertainty, context, and load conditions for users, services, and autonomous agents. The protocol is transport-independent and integrates with existing authentication mechanisms like OAuth and mTLS.\",\n \"n\": 4,\n \"nn\": \"Notable originality in introducing probabilistic authorization decisions - most authorization systems are deterministic, while this explicitly incorporates uncertainty and stochastic outcomes\",\n \"m\": 4,\n \"mn\": \"Detailed 44-page specification with defined protocol objects, decision semantics, HTTP binding, and examples - appears implementation-ready\",\n \"o\": 2,\n \"on\": \"Minor similarities to existing authorization frameworks but the stochastic decision-making approach is relatively unique in the authorization space\",\n \"mo\": 2,\n \"mon\": \"Single revision from November 2025 with no apparent working group adoption or community momentum yet\",\n \"r\": 4,\n \"rn\": \"Directly relevant to AI agents as it specifically addresses authorization for autonomous agents and incorporates uncertainty handling which is crucial for agent systems\",\n \"c\": [\"Agent identity/auth\", \"Policy/governance\"]\n}", "in_tok": 588, "out_tok": 280} {"draft_name": "draft-yang-ioa-protocol", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft defines the Internet of Agents Protocol (IoA) for enabling heterogeneous agent collaboration through dynamic team formation, adaptive coordination, and structured communication. The protocol uses a layered architecture with extensible messaging to support decentralized deployment across devices and interoperability with existing frameworks.\",\"n\":3,\"nn\":\"Useful contribution addressing multi-agent coordination, though builds on established distributed systems and agent collaboration concepts\",\"m\":2,\"mn\":\"Early sketch stage - defines general architecture and concepts but lacks detailed protocol specifications, message formats, and implementation details\",\"o\":3,\"on\":\"Shares concepts with existing multi-agent system protocols and distributed coordination mechanisms, but focuses specifically on heterogeneous agent collaboration\",\"mo\":2,\"mon\":\"Single revision draft with no clear working group adoption or community engagement visible\",\"r\":5,\"rn\":\"Core AI agent topic directly addressing fundamental challenge of multi-agent coordination and collaboration in distributed intelligent systems\",\"c\":[\"A2A protocols\",\"Agent discovery/reg\",\"Data formats/interop\"]}", "in_tok": 596, "out_tok": 218} {"draft_name": "draft-sovereign-svtp", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"SVTP proposes a protocol for establishing verifiable identity and trust for autonomous machines including AI agents and physical systems like robots and vehicles. It defines SVTP-DID identifiers and Protocol Seal mechanisms to enable secure machine-to-machine interactions with liability containment.\",\n \"n\": 4,\n \"nn\": \"Novel integration of DID-based identity with physical autonomous systems and liability containment - goes beyond typical identity frameworks\",\n \"m\": 2,\n \"mn\": \"Early specification stage with key concepts defined but lacking detailed protocol mechanics, message formats, and implementation guidance\",\n \"o\": 3,\n \"on\": \"Shares DID concepts with W3C standards and overlaps with some IoT trust frameworks, but unique autonomous machine focus\",\n \"mo\": 2,\n \"mon\": \"Single draft version with no indication of working group adoption or broader community engagement\",\n \"r\": 4,\n \"rn\": \"Directly addresses AI agent identity and autonomous system trust - core infrastructure need for agent ecosystems\",\n \"c\": [\"Agent identity/auth\", \"A2A protocols\", \"Policy/governance\"]\n}", "in_tok": 600, "out_tok": 265} {"draft_name": "draft-jennings-ai-mcp-over-moq", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft defines how to transport the Model Context Protocol (MCP) over Media over QUIC Transport (MOQT), enabling efficient delivery of AI agent capabilities including tools, resources, and composed \\\"Agent Skills\\\". The specification focuses on mapping MCP primitives to MOQT objects with appropriate prioritization and delivery guarantees for scalable AI agent deployments.\",\"n\":3,\"nn\":\"Useful contribution combining two emerging protocols (MCP and MOQT) in a natural way that leverages MOQT's streaming capabilities for AI agent communication\",\"m\":4,\"mn\":\"Detailed specification with clear mapping procedures, priority management, and relay architectures, though implementation examples could be stronger\",\"o\":2,\"on\":\"Minor overlap with general agent communication protocols but unique in its specific MCP-over-MOQT approach\",\"mo\":2,\"mon\":\"Single revision from 2026 with no clear working group adoption signals\",\"r\":5,\"rn\":\"Directly addresses core AI agent infrastructure needs for efficient communication and skill delivery at scale\",\"c\":[\"A2A protocols\",\"Data formats/interop\",\"Model serving/inference\"]}", "in_tok": 750, "out_tok": 248} {"draft_name": "draft-ramakrishna-satp-data-sharing", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft proposes a DLT-neutral protocol for requesting and sharing asset views across blockchain networks through gateway nodes. The protocol enables external agents to request views using global addresses while masking internal DLT complexities.\",\n \"n\": 3,\n \"nn\": \"Useful contribution addressing cross-DLT interoperability, though building on established gateway patterns\",\n \"m\": 3,\n \"mn\": \"Defines protocol mechanisms and view concepts but lacks detailed message formats and implementation specifics\",\n \"o\": 3,\n \"on\": \"Shares concepts with existing blockchain interoperability and gateway protocols but focuses specifically on view sharing\",\n \"mo\": 2,\n \"mon\": \"Single revision with unclear working group status or broader community engagement\",\n \"r\": 2,\n \"rn\": \"Tangentially related through external agents requesting views, but primarily focused on blockchain interoperability rather than AI agents\",\n \"c\": [\"Data formats/interop\", \"A2A protocols\"]\n}", "in_tok": 710, "out_tok": 237} {"draft_name": "draft-li-iotops-intelligent-security", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft proposes an intelligent protection optimization system for IoT networks using information-centric networking principles. The system aims to provide better content privacy protection through intelligent mechanisms in naming/parsing, data exchange, and data caching without additional costs.\",\n \"n\": 2,\n \"nn\": \"Incremental work combining existing ICN concepts with basic IoT security considerations, lacks novel technical insights\",\n \"m\": 2,\n \"mn\": \"Early sketch with high-level concepts but missing detailed protocol specifications, implementation details, and concrete mechanisms\",\n \"o\": 4,\n \"on\": \"Significant overlap with existing ICN security proposals and IoT protection schemes, particularly in caching and naming security\",\n \"mo\": 2,\n \"mon\": \"Single revision with minimal technical depth, no indication of working group interest or community engagement\",\n \"r\": 3,\n \"rn\": \"Partially relevant as it mentions intelligent protection systems but lacks clear AI/ML mechanisms or agent-based approaches\",\n \"c\": [\"Autonomous netops\", \"AI safety/alignment\"]\n}", "in_tok": 568, "out_tok": 252} {"draft_name": "draft-ietf-lamps-attestation-freshness", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft defines mechanisms for supplying nonces to end entities for freshness verification of attestation evidence in certificate signing requests across CMP, EST, and CMC protocols. It addresses the timing coordination needed between certificate authorities and attestation providers to ensure evidence freshness during certificate enrollment.\",\"n\":2,\"nn\":\"Incremental extension applying existing nonce-based freshness concepts to certificate management protocols\",\"m\":4,\"mn\":\"Detailed specification with concrete protocol extensions and examples for multiple certificate management frameworks\",\"o\":3,\"on\":\"Shares attestation freshness concepts with remote attestation standards but applies them specifically to certificate enrollment\",\"mo\":4,\"mon\":\"IETF LAMPS working group adoption with active development addressing real deployment needs\",\"r\":2,\"rn\":\"Tangentially related through attestation mechanisms but focused on certificate management rather than AI agents\",\"c\":[\"Agent identity/auth\"]}", "in_tok": 610, "out_tok": 193} {"draft_name": "draft-rosenberg-ai-protocols", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This framework document identifies communication protocols needed for AI agents to interact with users, APIs, and other agents over the Internet. It surveys existing work like MCP and A2A protocols while establishing requirements for potential IETF standardization in AI agent communications.\",\n \"n\": 2,\n \"nn\": \"Framework approach is incremental, cataloging existing protocols rather than proposing novel solutions\",\n \"m\": 2,\n \"mn\": \"Early framework sketch with use cases but lacks detailed protocol specifications or concrete mechanisms\",\n \"o\": 3,\n \"on\": \"Shares conceptual space with existing A2A and MCP work, though attempts to provide broader framework perspective\",\n \"mo\": 2,\n \"mon\": \"Single revision from individual contributor, no evidence of working group interest or broader community engagement\",\n \"r\": 5,\n \"rn\": \"Directly addresses core AI agent communication protocols and Internet-scale agent interactions\",\n \"c\": [\"A2A protocols\", \"Agent discovery/reg\", \"Data formats/interop\", \"Human-agent interaction\"]\n}", "in_tok": 697, "out_tok": 252} {"draft_name": "draft-ietf-rats-evidence-trans", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This IETF draft specifies methods for transforming different evidence formats (DiceTcbInfo, concise evidence, SPDM measurements) into a unified CoRIM representation for remote attestation procedures. It addresses the challenge of verifiers needing to handle multiple evidence encoding formats by providing standardized transformation methods.\",\"n\":2,\"nn\":\"Incremental work that applies existing CoRIM framework to additional evidence formats - useful but not groundbreaking\",\"m\":4,\"mn\":\"Detailed specification with concrete transformation methods for specific evidence structures, appears implementation-ready\",\"o\":3,\"on\":\"Builds directly on CoRIM and RATS architecture but focuses on specific transformation aspect not fully covered elsewhere\",\"mo\":4,\"mon\":\"IETF working group draft with clear progression and active development in established RATS framework\",\"r\":2,\"rn\":\"Remote attestation has some relevance to AI agent trust and verification but this is primarily a general security mechanism\",\"c\":[\"Agent identity/auth\"]}", "in_tok": 615, "out_tok": 218} {"draft_name": "draft-benzing-accp", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"ACCP defines a semantic encoding protocol for AI agent communication that claims 60-90% token reduction compared to natural language or JSON messaging. The protocol includes compact message formats, intent ontology, and state compression designed to complement existing agent protocols like MCP and A2A.\",\"n\":3,\"nn\":\"Useful contribution addressing real efficiency concerns in agent communication, though compression and semantic encoding are established techniques\",\"m\":3,\"mn\":\"Defines protocol structure and compression strategy but appears to lack detailed implementation specifics and test vectors\",\"o\":3,\"on\":\"Shares conceptual space with MCP and A2A protocols but focuses specifically on compression and efficiency rather than duplicating functionality\",\"mo\":2,\"mon\":\"Single revision with future date suggests early stage, no evidence of working group adoption or community engagement yet\",\"r\":5,\"rn\":\"Directly addresses core AI agent communication efficiency, highly relevant to current agent infrastructure development\",\"c\":[\"A2A protocols\",\"Data formats/interop\"]}", "in_tok": 558, "out_tok": 212} {"draft_name": "draft-sharif-attp-agent-trust-transport", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"ATTP defines a secure HTTP-based protocol for AI agent-to-server communication with mandatory cryptographic identity verification, digital signatures, and audit trails. The protocol adds security headers to HTTP requests/responses and requires Agent Passport JWT credentials for all interactions.\",\"n\":2,\"nn\":\"Incremental security extension to HTTP for agent communication, building on established JWT/ECDSA patterns\",\"m\":4,\"mn\":\"Detailed specification with defined headers, security requirements, and clear protocol mechanics\",\"o\":4,\"on\":\"Significant conceptual overlap with existing secure HTTP extensions and agent authentication schemes\",\"mo\":2,\"mon\":\"Single revision draft with no evident working group adoption or community feedback\",\"r\":4,\"rn\":\"Directly addresses AI agent security needs with practical protocol design for agent-server communication\",\"c\":[\"Agent identity/auth\",\"A2A protocols\"]}", "in_tok": 847, "out_tok": 187} {"draft_name": "draft-hartman-credential-broker-4-agents", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"CB4A specifies a credential brokering architecture for AI agents that eliminates long-lived credential storage by agents, instead providing short-lived proxy credentials through a policy-separated broker system. The draft builds on SPIFFE/SPIRE and DPoP to address credential sprawl risks in agentic AI deployments.\",\n \"n\": 4,\n \"nn\": \"Notable originality in applying credential brokering specifically to AI agent architectures with comprehensive threat modeling for agentic scenarios\",\n \"m\": 4,\n \"mn\": \"Detailed specification with three proxy models, tiered approval framework, and comprehensive threat analysis with mitigations\",\n \"o\": 2,\n \"on\": \"Minor similarities to general credential management approaches but unique focus on AI agent-specific challenges and architecture\",\n \"mo\": 2,\n \"mon\": \"Single author draft with no clear working group adoption signals, though addresses emerging AI agent security concerns\",\n \"r\": 5,\n \"rn\": \"Directly addresses core AI agent security challenge of credential management and sprawl prevention in multi-service agent deployments\",\n \"c\": [\"Agent identity/auth\", \"AI safety/alignment\", \"Policy/governance\"]\n}", "in_tok": 681, "out_tok": 282} {"draft_name": "draft-wolfe-faf-format", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft specifies FAF, a YAML-based format for representing software project context that AI development tools can consume. It aims to standardize how project architecture, conventions, and dependencies are shared across different AI agents in multi-agent development environments.\",\n \"n\": 2,\n \"nn\": \"Incremental work applying existing structured format (YAML) to AI tooling context - useful but not particularly novel\",\n \"m\": 3,\n \"mn\": \"Defines a specific format with IANA registration mentioned, suggesting concrete specification though implementation details may be limited\",\n \"o\": 3,\n \"on\": \"Shares concepts with other data format and interoperability efforts for AI systems, though specific to development tooling context\",\n \"mo\": 2,\n \"mon\": \"Single revision draft with IANA registration claim but no clear evidence of broader community engagement or working group adoption\",\n \"r\": 4,\n \"rn\": \"Directly addresses AI agent interoperability in software development, which is highly relevant to multi-agent systems and AI tooling\",\n \"c\": [\"Data formats/interop\", \"Human-agent interaction\"]\n}", "in_tok": 574, "out_tok": 270} {"draft_name": "draft-wmz-nmrg-agent-ndt-arch", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"Proposes an architecture combining Network Digital Twins with Agentic AI for intent-based network operations, allowing risk-free testing and optimization of network changes. Describes how existing technologies can be integrated to enable AI-driven autonomous network management.\",\n \"n\": 2,\n \"nn\": \"Incremental combination of existing NDT and Agentic AI concepts without novel architectural insights\",\n \"m\": 2,\n \"mn\": \"Early architectural sketch lacking detailed protocols, interfaces, or implementation specifics\",\n \"o\": 4,\n \"on\": \"Significant conceptual overlap with existing intent-based networking and digital twin literature\",\n \"mo\": 2,\n \"mon\": \"Single revision from research group with uncertain community traction\",\n \"r\": 4,\n \"rn\": \"Directly relevant to AI agents in networking with focus on autonomous operations and digital twins\",\n \"c\": [\"Autonomous netops\", \"Human-agent interaction\"]\n}", "in_tok": 722, "out_tok": 229} {"draft_name": "draft-berlinai-vera", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"VERA proposes a zero-trust architecture for AI agent security with five enforcement pillars, formal security properties, and cryptographic proof-based autonomy earning. The draft addresses critical risks from compromised AI agents including data exfiltration and unauthorized transactions.\",\"n\":4,\"nn\":\"Notable originality in combining zero-trust principles with agent-specific threat models and cryptographic proof-based autonomy earning\",\"m\":3,\"mn\":\"Defines structured architecture with typed schemas and formal properties, but lacks detailed implementation specifics\",\"o\":2,\"on\":\"Minor conceptual overlap with general zero-trust frameworks but novel application to AI agents\",\"mo\":2,\"mon\":\"Single revision with recent date, no indication of working group adoption yet\",\"r\":5,\"rn\":\"Core AI agent security topic addressing fundamental runtime enforcement challenges\",\"c\":[\"AI safety/alignment\",\"Agent identity/auth\",\"Policy/governance\"]}", "in_tok": 566, "out_tok": 194} {"draft_name": "draft-vasylenko-pait-protocol", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"PAIT defines a wire protocol for token-level provenance tracking in generative AI systems, including agent identity tokens, provenance manifests, and trust telemetry. The protocol aims to address AI transparency regulations by providing verifiable attribution for AI-generated content at sub-document granularity.\",\"n\":4,\"nn\":\"Notable originality in defining token-level provenance attribution with cryptographic binding to training corpus segments - addresses a specific gap in AI transparency not covered by existing protocols\",\"m\":3,\"mn\":\"Defines core protocol components and state machine but lacks detailed implementation specifics, message formats, and concrete examples needed for full implementation\",\"o\":2,\"on\":\"Minor conceptual similarities to general provenance and identity frameworks, but the token-level AI inference focus and specific regulatory compliance angle are distinct\",\"mo\":2,\"mon\":\"Single revision draft with no clear working group adoption or community feedback visible - appears to be early-stage individual contribution\",\"r\":5,\"rn\":\"Directly addresses core AI transparency and accountability challenges with practical protocol mechanisms for provenance tracking in generative AI systems\",\"c\":[\"Agent identity/auth\",\"Policy/governance\",\"Model serving/inference\",\"Data formats/interop\"]}", "in_tok": 725, "out_tok": 257} {"draft_name": "draft-aylward-aiga-1", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"Proposes AIGA protocol for governing autonomous AI agents through tiered risk-based oversight, immutable kernel architecture, and federated authority networks. Uses economic incentives, quarantine mechanisms, and multi-layered security for high-risk agent scenarios.\",\n \"n\": 4,\n \"nn\": \"Novel integration of TEE attestation, federated governance, and economic incentive alignment specifically for autonomous AI agents represents notable originality in the governance space\",\n \"m\": 3,\n \"mn\": \"Well-defined protocol architecture with clear components like Immutable Kernel and Federated Authority Network, though lacks detailed implementation specifics and test cases\",\n \"o\": 2,\n \"on\": \"While governance concepts exist broadly, the specific combination of AI agent oversight mechanisms and federated authority approach shows minimal overlap with existing IETF work\",\n \"mo\": 2,\n \"mn\": \"Single draft from 2026 with no clear working group adoption or community discussion visible, suggesting early-stage individual contribution\",\n \"r\": 5,\n \"rn\": \"Directly addresses core challenges in autonomous AI agent governance and accountability, which is central to safe AI deployment\",\n \"c\": [\"Policy/governance\", \"Agent identity/auth\", \"AI safety/alignment\"]\n}", "in_tok": 809, "out_tok": 296} {"draft_name": "draft-yu-agent-identifier-rdap", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft defines an RDAP profile for querying registration data about AI agent identifiers, allowing lookup of agent metadata like IPv6 locators, gateways, public keys, and operational state. It reuses the existing RDAP protocol structure but extends it with new object classes specific to agent identifier management.\",\"n\":3,\"nn\":\"Useful application of existing RDAP protocol to agent identifiers - not groundbreaking but addresses a real need for agent metadata lookup\",\"m\":4,\"mn\":\"Detailed specification with defined object classes, extensions, and query patterns - appears implementation-ready with good technical depth\",\"o\":2,\"on\":\"Builds on RDAP but applies it to a new domain - minimal overlap with existing agent-related work\",\"mo\":2,\"mon\":\"Single revision from 2026 (future date suggests early stage) - no clear WG adoption signals yet\",\"r\":4,\"rn\":\"Directly addresses agent infrastructure needs for identifier resolution and metadata access - core component for agent deployment\",\"c\":[\"Agent identity/auth\",\"Agent discovery/reg\",\"Data formats/interop\"]}", "in_tok": 686, "out_tok": 239} {"draft_name": "draft-zm-rtgwg-mcp-network-measurement", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft proposes using Model Context Protocol (MCP) to enable natural language-driven network measurement by treating network devices as MCP servers and controllers/management systems as MCP clients. The framework aims to provide AI-assisted network monitoring, fault diagnosis, and automated measurement workflows through standardized MCP communication.\",\n \"n\": 3,\n \"nn\": \"Useful application of MCP to network measurement domain, though conceptually straightforward adaptation of existing protocol\",\n \"m\": 2,\n \"mn\": \"Early stage draft with basic framework description but lacking detailed protocol specifications and implementation guidance\",\n \"o\": 2,\n \"on\": \"Some conceptual overlap with network automation approaches but MCP application to measurement appears relatively unique\",\n \"mo\": 2,\n \"mon\": \"Single draft submission with no evidence of working group adoption or broader community engagement\",\n \"r\": 4,\n \"rn\": \"Directly relevant as it applies AI agent protocols (MCP) to enable natural language network operations and intelligent automation\",\n \"c\": [\"A2A protocols\", \"Autonomous netops\", \"Human-agent interaction\"]\n}", "in_tok": 601, "out_tok": 262} {"draft_name": "draft-zeng-mcp-troubleshooting", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft proposes using the Model Context Protocol (MCP) to enable intent-based network troubleshooting by mapping network devices as MCP servers and controllers as MCP clients. It aims to provide conversational, automated network management through LLM integration without requiring new protocols.\",\n \"n\": 3,\n \"nn\": \"Useful adaptation of existing MCP protocol to network management domain, though the core concept of applying LLMs to network troubleshooting is well-established\",\n \"m\": 2,\n \"mn\": \"Early-stage proposal with basic problem statement and high-level mapping concepts, lacks detailed protocol specifications or implementation examples\",\n \"o\": 3,\n \"on\": \"Shares general concepts with other network automation and intent-based networking approaches, though specific MCP application appears unique\",\n \"mo\": 2,\n \"mon\": \"Single revision from October 2025, no evidence of working group adoption or broader community engagement yet\",\n \"r\": 4,\n \"rn\": \"Directly relevant to AI agents in networking, combining LLM capabilities with network operations through structured protocol interactions\",\n \"c\": [\"Autonomous netops\", \"Data formats/interop\", \"Human-agent interaction\"]\n}", "in_tok": 614, "out_tok": 286} {"draft_name": "draft-ni-a2a-ai-agent-security-requirements", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft outlines security requirements for AI agents across their operational lifecycle including provisioning, registration, discovery, and access control. It appears to be an early attempt at standardizing security frameworks for autonomous agent systems.\",\"n\":3,\"nn\":\"Addresses growing need for AI agent security standards, though security requirements analysis is a well-established approach\",\"m\":2,\"mn\":\"At 14 pages appears to be early stage requirements gathering rather than detailed specifications or protocols\",\"o\":2,\"on\":\"Security requirements for distributed systems exist but AI agent-specific security considerations are relatively unexplored in IETF\",\"mo\":2,\"mon\":\"Single draft submission with no clear working group adoption or community feedback visible\",\"r\":5,\"rn\":\"Directly addresses core AI agent security concerns across multiple operational phases - highly relevant to agent infrastructure\",\"c\":[\"Agent identity/auth\",\"Policy/governance\",\"Agent discovery/reg\",\"A2A protocols\"]}", "in_tok": 489, "out_tok": 200} {"draft_name": "draft-zhang-agent-gap-network", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft analyzes the problems and gaps in applying AI agents to mobile core networks. It identifies challenges in integrating autonomous agents into 5G/6G core network operations and management.\",\"n\":2,\"nn\":\"Incremental application of existing agent concepts to mobile networks\",\"m\":1,\"mn\":\"Only provides problem statement and gap analysis without proposed solutions\",\"o\":3,\"on\":\"Shares concepts with other network automation and AI-enabled networking drafts\",\"mo\":2,\"mon\":\"Single revision with no clear follow-up activity\",\"r\":4,\"rn\":\"Directly addresses AI agents in network infrastructure context\",\"c\":[\"Autonomous netops\",\"Other AI/agent\"]}", "in_tok": 466, "out_tok": 148} {"draft_name": "draft-srijal-agents-policy", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft proposes AGENTS.TXT, a plaintext policy file protocol that allows websites to specify access rules for automated clients, bots, and crawlers. It includes hash verification, directive specifications, and mandatory restrictive behavior for malformed files to prevent unintended access.\",\"n\":2,\"nn\":\"Incremental extension of robots.txt concept for broader automated clients\",\"m\":3,\"mn\":\"Defines protocol structure with directives and verification but lacks implementation details\",\"o\":4,\"on\":\"Significant conceptual overlap with robots.txt and existing web crawling standards\",\"mo\":2,\"mon\":\"Single revision draft with no clear working group involvement\",\"r\":4,\"rn\":\"Directly relevant to AI agents and automated systems accessing web resources\",\"c\":[\"Policy/governance\",\"A2A protocols\",\"Data formats/interop\"]}", "in_tok": 529, "out_tok": 182} {"draft_name": "draft-chen-ai-agent-auth-new-requirements", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft identifies authentication and authorization challenges posed by autonomous AI agents whose dynamic behavior and delegation relationships don't fit traditional identity-based security models. It proposes new requirements for managing dynamic agent behavior rather than just verifying static identities.\",\n \"n\": 4,\n \"nn\": \"Notable originality in recognizing that AI agent autonomy fundamentally breaks existing auth models designed for predictable human/software entities\",\n \"m\": 2,\n \"mn\": \"Early sketch that identifies problems and high-level requirements but lacks concrete protocols or detailed mechanisms\",\n \"o\": 2,\n \"on\": \"Minor similarities to general identity management work but addresses unique AI agent characteristics not covered elsewhere\",\n \"mo\": 2,\n \"mon\": \"Single revision with no evident working group activity or community engagement yet\",\n \"r\": 5,\n \"rn\": \"Core AI agent topic directly addressing fundamental security infrastructure needed for autonomous agent deployment\",\n \"c\": [\"Agent identity/auth\", \"Policy/governance\"]\n}", "in_tok": 557, "out_tok": 235} {"draft_name": "draft-verma-dmsc-nlip-notes", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft proposes using generative AI to create a flexible, natural language-based communication protocol that could replace the multitude of rigid APIs currently used for agent-to-agent interactions. The authors argue this approach would solve interoperability challenges and provide a single uniform API for agent communications including registration, discovery, telemetry and debugging.\",\"n\":4,\"nn\":\"Using natural language as a universal API abstraction layer for agent communication is notably original, though builds on established concepts of AI-mediated interfaces\",\"m\":1,\"mn\":\"Only presents the problem statement and high-level concept without any protocol definition, mechanisms, or technical details\",\"o\":2,\"on\":\"While agent communication is covered by other drafts, the natural language API approach appears largely unique in the IETF space\",\"mo\":2,\"mon\":\"Appears to be an early single submission without indication of working group interest or community engagement\",\"r\":5,\"rn\":\"Directly addresses core AI agent communication challenges and proposes AI-native solutions for agent interoperability\",\"c\":[\"A2A protocols\",\"Data formats/interop\",\"Agent discovery/reg\"]}", "in_tok": 605, "out_tok": 242} {"draft_name": "draft-ruan-spring-priority-flow-control-sid", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft proposes extending SRv6 behaviors to enable flow control in WANs for AI workloads using RDMA, specifically adapting datacenter PFC mechanisms for wide area networks. The approach leverages SRv6 segment lists to identify upstream nodes for congestion notification delivery.\",\"n\":3,\"nn\":\"Useful adaptation of datacenter flow control to WAN context for AI traffic, though builds heavily on existing PFC and SRv6 technologies\",\"m\":2,\"mn\":\"Early sketch with basic concept and example workflow, but lacks detailed protocol specification, message formats, and implementation details\",\"o\":3,\"on\":\"Shares concepts with existing SRv6 extensions and datacenter flow control mechanisms, but specific WAN application appears less explored\",\"mo\":2,\"mon\":\"Single revision from 2026 with no indication of working group activity or broader community engagement\",\"r\":4,\"rn\":\"Directly addresses AI infrastructure needs for distributed training and inference across datacenters, targeting specific RDMA over WAN requirements\",\"c\":[\"ML traffic mgmt\"]}", "in_tok": 896, "out_tok": 235} {"draft_name": "draft-kartha-internet20-ainative", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"Proposes Internet 2.0 as an AI-native web extension with HTTP+AI protocol, Model Resolution Network for AI service discovery, and AI-aware browsers. Introduces intent-based routing and treats AI models as first-class network entities with distributed discovery mechanisms.\",\"n\":4,\"nn\":\"Notable architectural vision combining AI-native protocols, intent-based routing, and distributed model discovery in a cohesive framework\",\"m\":2,\"mn\":\"Early conceptual sketch with basic component definitions but lacks detailed protocol specifications or implementation guidance\",\"o\":2,\"on\":\"Some conceptual overlap with AI service discovery and HTTP extensions but unique in comprehensive Internet 2.0 architectural approach\",\"mo\":2,\"mon\":\"Single revision draft with no clear working group activity or community engagement signals\",\"r\":5,\"rn\":\"Directly addresses core AI agent infrastructure including discovery, protocols, and distributed AI interaction mechanisms\",\"c\":[\"A2A protocols\",\"Agent discovery/reg\",\"Model serving/inference\"]}", "in_tok": 626, "out_tok": 212} {"draft_name": "draft-sogomonian-aiip-architecture", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"Defines AIIP, a dedicated protocol for autonomous AI systems to execute real-world actions through authenticated channels. Uses a resolve-invoke-execute-receipt pattern with cryptographic verification of execution outcomes.\",\"n\":4,\"nn\":\"Novel dedicated protocol architecture for AI agent execution with cryptographic receipts - goes beyond existing approaches by defining execution boundary\",\"m\":2,\"mn\":\"Early architectural sketch without detailed protocol mechanics, message formats, or concrete examples\",\"o\":2,\"on\":\"Some conceptual overlap with agent protocols but unique focus on execution plane and cryptographic verification\",\"mo\":2,\"mon\":\"Single draft submission with no visible community engagement or working group activity\",\"r\":5,\"rn\":\"Directly addresses core AI agent infrastructure - autonomous execution, verification, and real-world action delegation\",\"c\":[\"A2A protocols\",\"Agent identity/auth\"]}", "in_tok": 563, "out_tok": 182} {"draft_name": "draft-hood-agtp-merchant-identity", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft extends the Agent Transfer Protocol (AGTP) to address the merchant identity gap by defining merchant verification mechanisms parallel to existing agent identity protocols. It introduces Merchant Manifest Documents, trust tiers, and protocol integration points to enable mutual verification between agents and merchants in agentic commerce transactions.\",\n \"n\": 2,\n \"nn\": \"Incremental extension applying existing AGTP identity concepts to the merchant side, filling an obvious gap but using established patterns\",\n \"m\": 4,\n \"mn\": \"Detailed specification with concrete mechanisms, data structures, protocol integration points, and specific status codes, appears implementation-ready\",\n \"o\": 4,\n \"on\": \"Heavy structural dependence on base AGTP protocol and direct parallel to Agent Manifest concepts, essentially mirroring existing agent identity mechanisms\",\n \"mo\": 3,\n \"mon\": \"Active development evidenced by version progression (v00 to v01) with alignment to AGTP v05, but unclear broader community engagement\",\n \"r\": 5,\n \"rn\": \"Core AI agent infrastructure addressing fundamental trust and identity verification in autonomous commerce transactions\",\n \"c\": [\"A2A protocols\", \"Agent identity/auth\", \"Policy/governance\"]\n}", "in_tok": 884, "out_tok": 288} {"draft_name": "draft-klrc-aiagent-auth", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft proposes applying existing authentication standards like WIMSE and OAuth 2.0 to AI agent interactions rather than creating new protocols. It aims to provide a framework for agent authentication and authorization while identifying gaps for future standardization.\",\"n\":2,\"nn\":\"Incremental application of existing OAuth/WIMSE standards to AI agents without fundamental innovations\",\"m\":2,\"mn\":\"Early framework sketch that identifies approach but lacks detailed protocol specifications\",\"o\":4,\"on\":\"Significant conceptual overlap with existing OAuth, WIMSE, and general identity management work\",\"mo\":3,\"mon\":\"Active development indicated by recent 2026 date and practical focus on standards application\",\"r\":4,\"rn\":\"Directly addresses core AI agent authentication needs, highly relevant to agent deployment\",\"c\":[\"Agent identity/auth\"]}", "in_tok": 570, "out_tok": 177} {"draft_name": "draft-ietf-rats-ear", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"Defines EAT Attestation Result (EAR) message format for verifiers to encode appraisal results of attester evidence with trustworthiness vectors. Supports both simple and composite devices with multiple attesters, using CWT or JWT serialization.\",\n \"n\": 3,\n \"nn\": \"Useful standardization of attestation result format with trustworthiness vectors, building on existing EAT framework\",\n \"m\": 4,\n \"mn\": \"Detailed specification at 27 pages with defined message formats, trustworthiness vectors, and serialization options\",\n \"o\": 3,\n \"on\": \"Builds on EAT framework and shares concepts with other attestation/trust measurement standards\",\n \"mo\": 4,\n \"mon\": \"IETF working group draft with specific publication target date indicating active development\",\n \"r\": 3,\n \"rn\": \"Attestation and trust measurement relevant for AI agent verification and trustworthiness assessment\",\n \"c\": [\"Agent identity/auth\", \"Data formats/interop\"]\n}", "in_tok": 680, "out_tok": 259} {"draft_name": "draft-chang-agent-context-interaction", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft specifies procedures and optimizations for distributing context information among multiple agents in multi-agent systems to improve execution latency, reduce token consumption, and increase task completion rates. It focuses on precise control mechanisms for context distribution during complex multi-round workflows.\",\"n\":3,\"nn\":\"Addresses a legitimate technical challenge in multi-agent systems with context optimization focus\",\"m\":2,\"mn\":\"Early stage specification with basic problem definition but lacking detailed protocols\",\"o\":3,\"on\":\"Shares concepts with general multi-agent coordination work but specific context distribution angle is less common\",\"mo\":1,\"mon\":\"Single draft submission with no indication of community engagement or follow-up activity\",\"r\":5,\"rn\":\"Directly addresses core multi-agent system coordination challenges essential for AI agent deployments\",\"c\":[\"A2A protocols\",\"Data formats/interop\"]}", "in_tok": 531, "out_tok": 185} {"draft_name": "draft-loffredo-regext-rdap-verified-contacts", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft extends RDAP to include verification status information for contact fields like email addresses and phone numbers. It aims to improve data quality and trustworthiness by indicating which contact data has been verified and the verification methods used.\",\"n\":2,\"nn\":\"Incremental extension adding verification metadata to existing RDAP contact fields - useful but straightforward enhancement\",\"m\":4,\"mn\":\"Detailed specification with clear protocol extensions, field definitions, and comprehensive examples showing verified contact structures\",\"o\":3,\"on\":\"Shares verification concepts with other trust/authentication protocols but applies them specifically to RDAP contact data\",\"mo\":3,\"mon\":\"Active development as evidenced by 2026 date, but momentum unclear without seeing revision history or working group status\",\"r\":1,\"rn\":\"This is purely about domain registration data verification protocols, not related to AI agents or autonomous systems\",\"c\":[\"Other AI/agent\"]}", "in_tok": 532, "out_tok": 197} {"draft_name": "draft-narajala-ans", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft proposes Agent Name Service (ANS), a DNS-inspired directory system for AI agent discovery using PKI certificates for identity verification. It defines registration, renewal, and resolution mechanisms with a modular protocol adapter layer supporting various agent communication standards.\",\"n\":4,\"nn\":\"Novel application of DNS concepts to AI agent discovery with innovative capability-aware resolution and protocol-agnostic registry design\",\"m\":4,\"mn\":\"Detailed specification with JSON schemas, algorithms, threat analysis, and comprehensive protocol definitions approaching implementation readiness\",\"o\":2,\"on\":\"Unique focus on agent discovery with DNS-inspired approach, minimal overlap with existing IETF work\",\"mo\":2,\"mon\":\"Single draft with no apparent working group activity or community feedback yet\",\"r\":5,\"rn\":\"Directly addresses core AI agent infrastructure need for secure discovery and interoperability in multi-agent systems\",\"c\":[\"Agent discovery/reg\",\"Agent identity/auth\",\"A2A protocols\",\"Data formats/interop\"]}", "in_tok": 685, "out_tok": 215} {"draft_name": "draft-ietf-netconf-configuration-tracing", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft specifies a NETCONF mechanism to trace configuration changes back to their originating network management system using external trace IDs. It includes a YANG module to map local configuration changes to their sources, enabling better troubleshooting and root cause analysis in multi-NMS environments.\",\n \"n\": 2,\n \"nn\": \"Incremental extension of existing NETCONF capabilities with trace ID correlation - useful but straightforward addition to current practices\",\n \"m\": 4,\n \"mn\": \"Detailed specification with concrete YANG module and clear mechanisms for trace ID propagation and mapping across network management layers\",\n \"o\": 3,\n \"on\": \"Shares configuration management and tracing concepts with existing NETCONF/RESTCONF work, but specific external trace ID approach has limited direct overlap\",\n \"mo\": 4,\n \"mon\": \"IETF working group draft with recent revision date, indicating active development and institutional backing within established standards process\",\n \"r\": 4,\n \"rn\": \"Directly supports autonomous network operations by enabling automated root cause analysis and closed-loop automation for self-healing networks\",\n \"c\": [\"Autonomous netops\"]\n}", "in_tok": 713, "out_tok": 276} {"draft_name": "draft-mozley-aidiscovery", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft presents a problem statement for AI agent-to-agent discovery mechanisms, identifying requirements and considerations for how autonomous agents can find and connect with each other. It appears to be an early-stage exploration of discovery protocols needed as AI agents become more prevalent in networked environments.\",\n \"n\": 3,\n \"nn\": \"Addresses a legitimate emerging need for agent discovery protocols, building on established service discovery concepts but applying them to the AI agent domain\",\n \"m\": 1,\n \"mn\": \"Pure problem statement format with requirements gathering - no actual protocol mechanisms or technical solutions proposed\",\n \"o\": 3,\n \"on\": \"Shares conceptual ground with existing service discovery protocols (DNS-SD, mDNS) but focuses specifically on AI agent characteristics and requirements\",\n \"mo\": 2,\n \"mon\": \"Single draft submission with no clear working group adoption or broader community engagement visible\",\n \"r\": 5,\n \"rn\": \"Directly addresses core AI agent networking infrastructure needs - discovery is fundamental for agent-to-agent communication and coordination\",\n \"c\": [\"Agent discovery/reg\", \"A2A protocols\"]\n}", "in_tok": 484, "out_tok": 267} {"draft_name": "draft-kale-agntcy-federated-privacy", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft proposes a reference architecture for privacy-preserving federated learning across multi-tenant AI agent systems, combining federated averaging with differential privacy and secure aggregation. The approach aims to enable collaborative model training between organizations while maintaining tenant data isolation and formal privacy guarantees.\",\"n\":3,\"nn\":\"Useful contribution combining established techniques (federated learning, differential privacy) for the specific context of multi-tenant AI agents, but builds on well-known privacy-preserving ML approaches\",\"m\":2,\"mn\":\"Early sketch stage - only 13 pages for a complex architecture covering federated learning, differential privacy, and secure aggregation likely indicates high-level overview without detailed protocol specifications\",\"o\":3,\"on\":\"Shares significant conceptual overlap with existing federated learning and privacy-preserving ML literature, though specific application to multi-tenant AI agents may be less explored\",\"mo\":2,\"mon\":\"Single revision with recent 2026 date suggests early-stage work without demonstrated community engagement or working group adoption\",\"r\":4,\"rn\":\"Directly relevant to AI agent systems, addressing the important challenge of collaborative learning across organizational boundaries while preserving privacy in multi-tenant deployments\",\"c\":[\"AI safety/alignment\",\"Policy/governance\",\"Model serving/inference\"]}", "in_tok": 548, "out_tok": 271} {"draft_name": "draft-zeng-mcp-network-mgmt", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft extends the Model Context Protocol (MCP) to enable network equipment like routers and switches to act as MCP servers, allowing AI-driven network controllers to manage infrastructure through standardized JSON-RPC interactions. It defines new capability tokens, tools, resources, and error codes specifically for network management while maintaining backwards compatibility with existing MCP implementations.\",\"n\":3,\"nn\":\"Useful application of existing MCP framework to network management domain, though the core idea of AI-driven network operations is well-established\",\"m\":3,\"mn\":\"Defines protocol extensions and mechanisms but likely needs more detailed specifications and implementation examples for full deployment\",\"o\":2,\"on\":\"While network management protocols exist, applying MCP specifically to this domain has minimal direct overlap with other IETF work\",\"mo\":2,\"mon\":\"Single revision with October 2025 date, no clear indication of working group adoption or broader community engagement yet\",\"r\":4,\"rn\":\"Directly relevant to AI agents in networking, addressing practical integration between AI systems and network infrastructure management\",\"c\":[\"Autonomous netops\",\"A2A protocols\",\"Data formats/interop\"]}", "in_tok": 548, "out_tok": 245} {"draft_name": "draft-williams-http-bearer-extension", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft proposes extensions to HTTP Bearer authentication to enable automatic token fetching from Security Token Services during 401/407 flows. It aims to improve user experience by eliminating PKCE requirements and allowing token caching while providing OIDC fallback mechanisms.\",\"n\":2,\"nn\":\"Incremental improvement to existing Bearer auth flows, combining well-known patterns but not introducing fundamentally new concepts\",\"m\":3,\"mn\":\"Defines specific protocol mechanisms and flows with clear examples, though could benefit from more implementation details\",\"o\":4,\"on\":\"Significant overlap with existing OAuth 2.0, OIDC, and HTTP authentication RFCs, building incrementally on established standards\",\"mo\":2,\"mon\":\"Appears to be single revision with limited evidence of broader community engagement or working group adoption\",\"r\":3,\"rn\":\"Partially relevant as improved authentication flows could benefit AI agents and automated systems, though not specifically targeting AI use cases\",\"c\":[\"Agent identity/auth\",\"A2A protocols\"]}", "in_tok": 578, "out_tok": 220} {"draft_name": "draft-schulze-ecap", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"ECAP proposes a cryptographically-secured protocol to replace robots.txt, requiring verifiable consent agreements between websites and AI crawlers/scrapers. The protocol uses digital signatures and declarative policies to create enforceable access controls rather than voluntary guidelines.\",\n \"n\": 4,\n \"nn\": \"Notable originality in applying cryptographic verification to web crawling consent - moves beyond voluntary standards to enforceable agreements\",\n \"m\": 2,\n \"mn\": \"Early sketch stage - only 3 pages with abstract-level description, lacks protocol details, message formats, or implementation guidance\",\n \"o\": 2,\n \"on\": \"Minor conceptual similarities to existing authentication protocols but unique application to crawler consent management\",\n \"mo\": 2,\n \"mon\": \"Single revision with recent date suggests initial proposal without demonstrated community engagement or working group adoption\",\n \"r\": 5,\n \"rn\": \"Core AI agent topic - directly addresses critical issue of AI scraper consent and ethical data collection that's highly relevant to current AI deployment concerns\",\n \"c\": [\"Policy/governance\", \"Agent identity/auth\", \"A2A protocols\"]\n}", "in_tok": 534, "out_tok": 270} {"draft_name": "draft-zheng-dispatch-agent-identity-management", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft specifies agent identity management for an Internet of Agents (IOA) system, defining requirements for agent identities, registration processes, and identifier structures. It outlines basic and extended identity management functions to be performed by agent gateways.\",\n \"n\": 2,\n \"nn\": \"Incremental work applying traditional identity management concepts to agent systems without significant novel approaches\",\n \"m\": 2,\n \"mn\": \"Early sketch level - provides high-level framework but lacks detailed protocol specifications and implementation details\",\n \"o\": 4,\n \"on\": \"Significant conceptual overlap with existing identity management systems and other agent infrastructure proposals\",\n \"mo\": 2,\n \"mon\": \"Single revision with no apparent working group activity or broader community engagement\",\n \"r\": 5,\n \"rn\": \"Core AI agent topic directly addressing fundamental infrastructure needs for agent identification and management\",\n \"c\": [\"Agent identity/auth\", \"Agent discovery/reg\", \"A2A protocols\"]\n}", "in_tok": 521, "out_tok": 236} {"draft_name": "draft-zzn-dvs", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"Defines Domain-Verified Skills (DVS) protocol for AI agents to discover and execute skill definitions served over HTTPS, using domain-based trust without centralized registries. Trust is established through Trust Root URL prefixes that scope verification boundaries.\",\n \"n\": 4,\n \"nn\": \"Novel application of DNS/TLS trust infrastructure to AI agent skill verification, creative path-scoping solution for UGC platforms\",\n \"m\": 4,\n \"mn\": \"Well-defined protocol with clear trust model, specific mechanisms for verification, and practical considerations for different hosting scenarios\",\n \"o\": 2,\n \"on\": \"Some conceptual similarity to general service discovery but unique AI agent skill focus and domain-based trust approach\",\n \"mo\": 2,\n \"mn\": \"Single draft submission with no apparent working group adoption or community discussion yet\",\n \"r\": 5,\n \"rn\": \"Directly addresses core AI agent needs for skill discovery, verification, and execution with practical trust model\",\n \"c\": [\"Agent discovery/reg\", \"Agent identity/auth\", \"Data formats/interop\"]\n}", "in_tok": 787, "out_tok": 262} {"draft_name": "draft-gaikwad-llm-benchmarking-terminology", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft establishes standardized terminology for benchmarking Large Language Model inference serving systems, covering latency, throughput, resource utilization, and quality metrics. It provides a shared vocabulary for evaluating inference engines, application gateways, and compound agentic systems without prescribing specific methodologies.\",\"n\":2,\"nn\":\"Incremental work defining terminology for an emerging but well-understood domain\",\"m\":4,\"mn\":\"Detailed terminology definitions with clear scope and comprehensive coverage of LLM serving metrics\",\"o\":2,\"on\":\"Limited overlap as LLM serving benchmarking terminology is relatively specialized\",\"mo\":3,\"mon\":\"Active development in emerging field but no clear WG adoption signals\",\"r\":4,\"rn\":\"Directly relevant to AI agent infrastructure as LLM serving is core to agent systems\",\"c\":[\"Model serving/inference\"]}", "in_tok": 546, "out_tok": 188} {"draft_name": "draft-dunbar-onsen-ac-pe2pe-ucmp-applicability", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft explores applying existing IETF YANG data models (AC, TE Topology, SR Policy, QoS) to support dynamic UCMP policy enforcement for AI inference workloads across distributed edge cloud sites. It serves as a practical evaluation of current YANG model applicability and identifies gaps in cloud-initiated, time-bounded network policy coordination.\",\n \"n\": 2,\n \"nn\": \"Incremental application of existing YANG models to AI inference use case, with straightforward mapping exercise\",\n \"m\": 2,\n \"mn\": \"Early sketch exploring applicability with gap identification but lacks detailed protocol mechanisms or implementation specifics\",\n \"o\": 3,\n \"on\": \"Shares networking automation concepts with existing TE and QoS drafts, though specific AI inference application provides some differentiation\",\n \"mo\": 2,\n \"mon\": \"Single revision with niche focus on YANG model applicability; limited broader community engagement evident\",\n \"r\": 3,\n \"rn\": \"Partially relevant - addresses network infrastructure requirements for AI inference workloads but focuses more on networking protocols than AI agent behavior\",\n \"c\": [\"ML traffic mgmt\", \"Autonomous netops\"]\n}", "in_tok": 677, "out_tok": 285} {"draft_name": "draft-ietf-tls-ecdhe-mlkem", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft specifies hybrid post-quantum key agreement mechanisms for TLS 1.3 that combine ML-KEM lattice-based cryptography with traditional ECDHE. It defines three specific combinations to provide quantum-resistant security while maintaining compatibility with current elliptic curve implementations.\",\n \"n\": 2,\n \"nn\": \"Incremental work combining well-established post-quantum ML-KEM with standard ECDHE - straightforward hybrid approach following expected migration patterns\",\n \"m\": 4,\n \"mn\": \"Detailed specification with concrete parameter combinations, algorithm definitions, and appears implementation-ready for TLS 1.3 integration\",\n \"o\": 4,\n \"on\": \"Significant overlap with other post-quantum TLS drafts and hybrid key exchange mechanisms - builds on established ML-KEM and ECDHE standards\",\n \"mo\": 5,\n \"mon\": \"IETF working group draft with strong industry momentum for post-quantum cryptography migration, critical for quantum-resistant TLS deployment\",\n \"r\": 1,\n \"rn\": \"Pure cryptographic protocol specification for general TLS security - not specifically related to AI agents or autonomous systems\",\n \"c\": [\"Other AI/agent\"]\n}", "in_tok": 568, "out_tok": 292} {"draft_name": "draft-ietf-lake-ra", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft specifies how to integrate remote attestation capabilities into the EDHOC lightweight key exchange protocol using the RATS architecture. It enables verification of device/software integrity during the authentication handshake process.\",\n \"n\": 3,\n \"nn\": \"Useful integration of existing RATS attestation with EDHOC, providing practical value for IoT security scenarios\",\n \"m\": 4,\n \"mn\": \"Detailed specification with protocol flows and message formats, appears implementation-ready\",\n \"o\": 2,\n \"on\": \"Builds on established EDHOC and RATS frameworks but creates novel integration approach\",\n \"mo\": 4,\n \"mon\": \"IETF working group draft with clear adoption path and active development\",\n \"r\": 1,\n \"rn\": \"This is about cryptographic protocols and device attestation, not AI agents or artificial intelligence systems\",\n \"c\": [\"Other AI/agent\"]\n}", "in_tok": 512, "out_tok": 228} {"draft_name": "draft-zhang-aiproto-svcb-mapping-for-agents", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft proposes a new DNS resource record type called AGENT that extends SVCB for discovering and resolving AI agent services. It addresses limitations in current DNS for agent service resolution by defining SVCB-compatible mappings specifically for intelligent agent communication protocols.\",\"n\":3,\"nn\":\"Useful contribution applying existing SVCB framework to emerging agent discovery needs\",\"m\":2,\"mn\":\"Early sketch with basic problem statement and proposed approach but lacks detailed specification\",\"o\":2,\"on\":\"Minor similarities to SVCB extensions but addresses distinct agent-specific use case\",\"mo\":1,\"mon\":\"Single draft with no apparent working group activity or community engagement\",\"r\":4,\"rn\":\"Directly addresses agent infrastructure needs for service discovery and protocol negotiation\",\"c\":[\"Agent discovery/reg\",\"A2A protocols\",\"Data formats/interop\"]}", "in_tok": 519, "out_tok": 186} {"draft_name": "draft-yang-alto-multi-domain", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft identifies use cases and challenges for ALTO (Application-Layer Traffic Optimization) in multi-domain network environments where traffic paths span multiple autonomous networks. It outlines three specific multi-domain scenarios and briefly discusses implementation approaches to address the identified challenges.\",\"n\":2,\"nn\":\"Incremental extension of existing ALTO work to multi-domain scenarios, fairly straightforward problem identification\",\"m\":2,\"mn\":\"Early stage work presenting use cases and challenges with only brief mentions of implementation solutions\",\"o\":3,\"on\":\"Builds directly on existing ALTO specifications but addresses a distinct multi-domain aspect\",\"mo\":2,\"mon\":\"Single revision draft from January 2024 with no clear indicators of working group adoption or community engagement\",\"r\":2,\"rn\":\"While about traffic optimization which could involve AI applications, this is primarily a networking protocol extension without direct AI/agent focus\",\"c\":[\"ML traffic mgmt\"]}", "in_tok": 583, "out_tok": 198} {"draft_name": "draft-irtf-nmrg-llm-nm", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft proposes a framework for integrating Large Language Models into network management systems with mandatory human oversight and decision approval loops. It defines telemetry modules, LLM decision components, and standardized data models to enable human-AI collaboration in network operations.\",\n \"n\": 3,\n \"nn\": \"Useful contribution combining LLMs with network management, though human-in-the-loop concepts are established in other domains\",\n \"m\": 2,\n \"mn\": \"Early framework sketch with high-level components described but lacking detailed protocols, APIs, or implementation specifics\",\n \"o\": 2,\n \"on\": \"Some conceptual overlap with general AI-assisted operations frameworks but specific network management focus is relatively unique\",\n \"mo\": 3,\n \"mon\": \"Active development in IRTF research group with growing interest in AI-driven network operations, though still early stage\",\n \"r\": 5,\n \"rn\": \"Core AI agent topic directly addressing human-AI collaboration in critical infrastructure management\",\n \"c\": [\"Human-agent interaction\", \"Autonomous netops\", \"Data formats/interop\"]\n}", "in_tok": 566, "out_tok": 266} {"draft_name": "draft-song-anp-aip", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft proposes Agent Internet Protocol (AIP) as a network-layer protocol for AI agent communication, using agent:// URIs instead of IP addresses and serving as the narrow waist for an Agent Network Protocol suite. It defines message formats, addressing, and processing rules analogous to IP but tailored for autonomous AI agents.\",\"n\":4,\"nn\":\"Creating a dedicated network protocol layer specifically for AI agents with human-readable capability-descriptive naming is notably original\",\"m\":4,\"mn\":\"Comprehensive 32-page specification with detailed message formats, addressing model, error handling, and layer interfaces suggests implementation-ready detail\",\"o\":2,\"on\":\"While building on IP concepts, the agent-specific addressing and protocol design appears largely unique\",\"mo\":2,\"mon\":\"Single draft from 2026 with no indication of working group adoption or broader community engagement\",\"r\":5,\"rn\":\"Directly addresses core AI agent networking infrastructure, proposing fundamental protocols for agent-to-agent communication\",\"c\":[\"A2A protocols\",\"Agent identity/auth\",\"Agent discovery/reg\"]}", "in_tok": 660, "out_tok": 233} {"draft_name": "draft-deshpande-rats-multi-verifier", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft extends IETF RATS architecture to support multiple verifiers coordinating together to assess attester trustworthiness. It defines topological patterns for multi-verifier systems while remaining agnostic to specific protocols or implementations.\",\"n\":3,\"nn\":\"Useful architectural extension addressing real need for distributed attestation in complex systems\",\"m\":2,\"mn\":\"Early architectural sketch with basic patterns but lacks detailed mechanisms or protocols\",\"o\":2,\"on\":\"Minor overlap with base RATS architecture but addresses distinct multi-verifier coordination gap\",\"mo\":2,\"mon\":\"Single revision with February 2026 date, appears to be early-stage work\",\"r\":2,\"rn\":\"Remote attestation is tangentially related to AI agents but primarily focused on general system trustworthiness\",\"c\":[\"Agent identity/auth\"]}", "in_tok": 566, "out_tok": 180} {"draft_name": "draft-aylward-daap-v2", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"DAAP v2.0 provides a comprehensive protocol for monitoring and controlling autonomous AI agents through cryptographic identity verification, periodic check-ins, and remote shutdown capabilities. The protocol addresses AI safety concerns with hardware-backed security, real-time behavioral monitoring, and ethical governance frameworks.\",\n \"n\": 4,\n \"nn\": \"Notable originality in combining hardware-backed security with AI-specific behavioral monitoring and remote shutdown capabilities in a standardized protocol framework\",\n \"m\": 4,\n \"mn\": \"Detailed 18-page specification with comprehensive protocol mechanisms including authentication, monitoring, and control systems\",\n \"o\": 2,\n \"on\": \"Minor similarities to general device management protocols but unique application to AI agent accountability and safety\",\n \"mo\": 2,\n \"mon\": \"Single revision with no indication of working group adoption or broader community engagement\",\n \"r\": 5,\n \"rn\": \"Core AI agent topic directly addressing critical challenges in AI safety, accountability, and autonomous agent control\",\n \"c\": [\"AI safety/alignment\", \"Agent identity/auth\", \"Policy/governance\", \"A2A protocols\"]\n}", "in_tok": 579, "out_tok": 265} {"draft_name": "draft-liang-agentdns", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"AgentDNS proposes a DNS-inspired naming and service discovery system that enables LLM agents to autonomously find and securely invoke third-party agent services across different vendors. The system introduces a unified namespace with semantic discovery, protocol-aware interoperability, and unified authentication for large-scale agent collaboration.\",\n \"n\": 4,\n \"nn\": \"Novel application of DNS concepts to agent ecosystems with semantic discovery and cross-vendor interoperability - addresses emerging need for agent-to-agent service discovery\",\n \"m\": 2,\n \"mn\": \"Early conceptual draft at 22 pages but likely lacks detailed protocol specifications, message formats, and implementation details needed for a working system\",\n \"o\": 2,\n \"on\": \"While inspired by DNS, the agent-specific semantic discovery and cross-vendor authentication aspects appear relatively unique in the agent networking space\",\n \"mo\": 2,\n \"mon\": \"Single draft from 2026 with no apparent working group adoption or community development indicated\",\n \"r\": 5,\n \"rn\": \"Directly addresses core agent infrastructure needs for service discovery, interoperability, and trust management in multi-agent systems\",\n \"c\": [\"Agent discovery/reg\", \"Agent identity/auth\", \"A2A protocols\", \"Data formats/interop\"]\n}", "in_tok": 593, "out_tok": 307} {"draft_name": "draft-howe-sipcore-mcp-extension", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft extends SIP to support the Model Context Protocol (MCP), adding option tags, headers, and media types for AI model context exchange. It enables MCP payload transmission during SIP session establishment and mid-dialog communications.\",\"n\":3,\"nn\":\"Useful integration of existing protocols (SIP + MCP) for a legitimate use case, though the extension follows standard SIP patterns\",\"m\":4,\"mn\":\"Detailed specification with new option tags, headers, media types, and multiple exchange mechanisms - appears implementation-ready\",\"o\":2,\"on\":\"While building on established SIP extension patterns, the specific MCP integration appears to be a unique approach\",\"mo\":2,\"mon\":\"Single revision from April 2026 with no visible working group activity or community discussion\",\"r\":4,\"rn\":\"Directly relevant to AI agent communication infrastructure, enabling context sharing between agents via established telephony protocols\",\"c\":[\"A2A protocols\",\"Data formats/interop\",\"Agent discovery/reg\"]}", "in_tok": 590, "out_tok": 215} {"draft_name": "draft-spm-lake-pqsuites", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft specifies quantum-resistant cipher suites for the EDHOC lightweight authenticated key exchange protocol, integrating ML-DSA signatures and ML-KEM key exchange algorithms. It defines how EDHOC operates with post-quantum cryptography for both signature-based and PSK-based authentication methods.\",\n \"n\": 2,\n \"nn\": \"Straightforward integration of standardized post-quantum algorithms into existing EDHOC protocol - incremental but necessary work\",\n \"m\": 4,\n \"mn\": \"Detailed specification with concrete cipher suite definitions and protocol operations, appears implementation-ready\",\n \"o\": 3,\n \"on\": \"Shares post-quantum cryptography integration concepts with other protocol upgrade drafts, but specific to EDHOC\",\n \"mo\": 4,\n \"mon\": \"Strong momentum given NIST PQC standardization completion and industry urgency for quantum-resistant protocols\",\n \"r\": 1,\n \"rn\": \"Pure cryptographic protocol specification focused on quantum resistance - not related to AI or autonomous agents\",\n \"c\": [\"Other AI/agent\"]\n}", "in_tok": 571, "out_tok": 265} {"draft_name": "draft-xsaopig-nsttlp-traffic-labeling", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"Proposes a protocol for embedding service type identifiers into network packets to enable intelligent traffic recognition and policy-based forwarding. The mechanism works across IPv4/IPv6 headers, MPLS labels, and Ethernet frames for applications like VR, scientific computing, and IoT.\",\"n\":2,\"nn\":\"Incremental extension of existing traffic classification and QoS marking concepts to include service type awareness\",\"m\":3,\"mn\":\"Defines a protocol mechanism with header formats but lacks detailed implementation specifics and test cases\",\"o\":4,\"on\":\"Significant overlap with existing QoS marking (DSCP), traffic engineering, and service chaining protocols\",\"mo\":2,\"mon\":\"Single draft with fabricated working group name suggests limited real development activity\",\"r\":2,\"rn\":\"Only tangentially AI-related through mention of 'intelligent' traffic recognition, primarily a network QoS protocol\",\"c\":[\"ML traffic mgmt\"]}", "in_tok": 595, "out_tok": 206} {"draft_name": "draft-rosenberg-aiproto", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft defines N-ACT, a standardized protocol for AI agents to discover and invoke third-party tools via normalized APIs. It focuses on enterprise environments and defines enumeration APIs for tool discovery and invocation APIs for runtime execution.\",\n \"n\": 2,\n \"nn\": \"Incremental standardization of existing AI tool integration patterns, building on established OpenAI function calling approaches\",\n \"m\": 3,\n \"mn\": \"Defines core protocol mechanisms with API specifications, though lacks detailed implementation examples and comprehensive error handling\",\n \"o\": 4,\n \"on\": \"Significant conceptual overlap with existing tool integration frameworks and OpenAI function calling specifications\",\n \"mo\": 2,\n \"mon\": \"Single draft submission with no evidence of working group adoption or broader community engagement\",\n \"r\": 5,\n \"rn\": \"Directly addresses core AI agent functionality around tool integration, a fundamental capability for enterprise agent deployment\",\n \"c\": [\"A2A protocols\", \"Data formats/interop\", \"Agent discovery/reg\"]\n}", "in_tok": 719, "out_tok": 248} {"draft_name": "draft-song-anp-ans", "model": "claude-sonnet-4-20250514", "raw": "```json\n{\n \"s\": \"Defines Agent Name System (ANS) for mapping human-readable agent:// URIs to cryptographic peer identities in autonomous AI agent networks. Provides name registration/resolution protocol with gossip and DHT dissemination mechanisms.\",\n \"n\": 4,\n \"nn\": \"Novel application of naming systems specifically for AI agent networks with agent:// URI scheme and multi-modal addressing (unicast/anycast/channel)\",\n \"m\": 4,\n \"mn\": \"Detailed 34-page specification with defined record formats, AITP methods, resolution algorithms, and dissemination mechanisms\",\n \"o\": 2,\n \"on\": \"Some conceptual similarity to DNS and other naming systems but specifically designed for AI agent peer-to-peer networks\",\n \"mo\": 2,\n \"mon\": \"Single draft from March 2026 with no clear working group adoption or community feedback visible\",\n \"r\": 5,\n \"rn\": \"Core AI agent infrastructure addressing fundamental naming and identity binding requirements for autonomous agent networks\",\n \"c\": [\"Agent identity/auth\", \"Agent discovery/reg\", \"A2A protocols\"]\n}\n```", "in_tok": 687, "out_tok": 274} {"draft_name": "draft-hood-agtp-api", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft specifies AGTP-API, a contract layer for the Agent Transfer Protocol that defines method catalogs, path grammar, endpoints, and runtime negotiation mechanisms for autonomous agent interactions. It consolidates and supersedes previous AGIS and AGTP-Methods drafts into a unified API specification.\",\n \"n\": 4,\n \"nn\": \"Notable originality in defining a comprehensive API contract layer specifically for autonomous agents, including novel concepts like semantic blocks, synthesis runtime negotiation, and structural rejection codes\",\n \"m\": 4,\n \"mn\": \"Detailed specification with comprehensive coverage of API components, method catalogs, endpoint definitions, and specific status codes, appears implementation-ready\",\n \"o\": 2,\n \"on\": \"Minor similarities to standard REST API design patterns, but the autonomous agent context and specific mechanisms like PROPOSE-and-synthesis are fairly unique\",\n \"mo\": 3,\n \"mon\": \"Active development evidenced by consolidation of previous drafts and comprehensive rewrite, but unclear if there's broader community adoption beyond the author\",\n \"r\": 5,\n \"rn\": \"Core AI agent topic focusing specifically on autonomous agent communication protocols and API design\",\n \"c\": [\"A2A protocols\", \"Data formats/interop\", \"Agent discovery/reg\"]\n}", "in_tok": 740, "out_tok": 300} {"draft_name": "draft-bernardos-nmrg-agentic-network-optimization", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft proposes using agentic AI for distributed sensing and communication optimization in ISAC networks, addressing the complex joint selection of sensing entities, processing locations, and network configurations. It presents initial architectural considerations for enabling closed-loop reconfiguration of sensing and networking resources under time-varying conditions.\",\n \"n\": 3,\n \"nn\": \"Useful contribution applying agentic AI to ISAC networks, though builds incrementally on existing sensing and AI concepts\",\n \"m\": 2,\n \"mn\": \"Early sketch with initial considerations and potential solution directions, lacks detailed mechanisms or protocols\",\n \"o\": 2,\n \"on\": \"Minor similarities to other network optimization drafts but focuses on unique ISAC sensing scenario\",\n \"mo\": 2,\n \"mon\": \"Single revision draft with future date, no evidence of working group adoption or community engagement yet\",\n \"r\": 4,\n \"rn\": \"Directly relevant as it specifically addresses agentic AI for network optimization with distributed decision-making\",\n \"c\": [\"Autonomous netops\", \"ML traffic mgmt\"]\n}", "in_tok": 822, "out_tok": 258} {"draft_name": "draft-li-cats-aisemantic-contract", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft defines a semantic-driven traffic shaping contract that allows AI applications to provide network metadata about tensor flows to enable differentiated forwarding and resource allocation. The approach aims to improve efficiency by moving beyond treating AI traffic as opaque byte streams to semantic-aware network scheduling.\",\n \"n\": 4,\n \"nn\": \"Notable originality in bridging AI application semantics with network-layer traffic management through explicit contracts\",\n \"m\": 2,\n \"mn\": \"Early sketch stage - defines concept but lacks detailed protocol mechanisms, message formats, or implementation specifics\",\n \"o\": 2,\n \"on\": \"Minor similarities to general QoS and traffic engineering work, but semantic-driven AI traffic contracts appear relatively unique\",\n \"mo\": 2,\n \"mon\": \"Single revision with no clear working group adoption or broader community engagement visible\",\n \"r\": 5,\n \"rn\": \"Core AI agent topic - directly addresses network infrastructure optimization for AI training and inference workloads\",\n \"c\": [\"ML traffic mgmt\", \"Autonomous netops\"]\n}", "in_tok": 609, "out_tok": 252} {"draft_name": "draft-hood-independent-agtp", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"AGTP defines a dedicated application-layer protocol for AI agent traffic that addresses limitations of HTTP for agent communication. It features runtime contract negotiation with a 12-method floor organized as cognitive and mechanics verbs, plus agent identity addressing.\",\n \"n\": 4,\n \"nn\": \"Creating a dedicated transport protocol for AI agents with runtime contract negotiation is notably original compared to existing HTTP-based approaches\",\n \"m\": 4,\n \"mn\": \"Very detailed 133-page specification with IANA registrations, formal URI grammar, complete method definitions, and comprehensive protocol mechanics\",\n \"o\": 2,\n \"on\": \"While it mentions existing AGMPs (MCP, ACP, A2A, ANP), AGTP positions itself as a transport substrate rather than competing messaging protocol\",\n \"mo\": 3,\n \"mon\": \"Version 07 indicates active development with significant changes, but unclear if this has broader community adoption or WG interest\",\n \"r\": 5,\n \"rn\": \"Directly addresses core AI agent infrastructure needs including identity, observability, and semantic vocabulary for agent-to-agent communication\",\n \"c\": [\"A2A protocols\", \"Agent identity/auth\", \"Agent discovery/reg\"]\n}", "in_tok": 959, "out_tok": 291} {"draft_name": "draft-annevk-johannhof-httpbis-cookies", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft updates the HTTP Cookie and Set-Cookie header field specifications, replacing RFC 6265 to address historical security and privacy issues while maintaining compatibility with existing widespread usage. The document provides a comprehensive revision of the cookie mechanism that underlies stateful HTTP sessions across the web.\",\"n\":2,\"nn\":\"Incremental improvement over existing cookie standards, addressing known issues rather than introducing novel concepts\",\"m\":4,\"mn\":\"Detailed specification with comprehensive coverage of cookie mechanisms, ready for implementation\",\"o\":5,\"on\":\"Direct update/replacement of RFC 6265 and 6265bis, essentially the same core functionality\",\"mo\":4,\"mon\":\"Active HTTPBIS working group development with clear path to standardization\",\"r\":1,\"rn\":\"HTTP cookie mechanism is fundamental web infrastructure, not specifically related to AI or agents\",\"c\":[\"Other AI/agent\"]}", "in_tok": 560, "out_tok": 188} {"draft_name": "draft-wang-cats-innetwork-infer", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft proposes a framework for decentralized LLM inference across untrusted participants, treating inference as distributed layer-wise execution with explicit deadlines rather than monolithic computation. It addresses the limitations of centralized LLM deployments by enabling coordination among heterogeneous compute resources without requiring trusted operators.\",\n \"n\": 4,\n \"nn\": \"Notable originality in applying layer-aware distributed execution to LLM inference with explicit deadline constraints and untrusted participant coordination - goes beyond typical federated learning approaches\",\n \"m\": 2,\n \"mn\": \"Early sketch phase - explicitly states it does not specify concrete protocols, wire formats, or state machines, focusing only on architectural framework and problem definition\",\n \"o\": 2,\n \"on\": \"Minor similarities to general distributed computing and federated learning concepts, but the LLM-specific layer-aware approach with deadline constraints appears relatively unique\",\n \"mo\": 2,\n \"mn\": \"Single revision with March 2026 date suggests very early stage work without established community engagement or working group adoption\",\n \"r\": 5,\n \"rn\": \"Core AI agent topic directly addressing LLM inference infrastructure, which is fundamental to AI agent deployment and operation at scale\",\n \"c\": [\"Model serving/inference\", \"A2A protocols\", \"Autonomous netops\"]\n}", "in_tok": 792, "out_tok": 310} {"draft_name": "draft-pocero-authkem-ikr-edhoc", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft specifies an optimized KEM-based authentication method for EDHOC protocol in scenarios where the initiator knows the responder's credentials beforehand. The approach enables post-quantum signature-free authentication using a three-message handshake while maintaining mutual authentication and forward secrecy.\",\n \"n\": 2,\n \"nn\": \"Incremental optimization of existing KEM-based EDHOC authentication for a specific but common use case\",\n \"m\": 4,\n \"mn\": \"Detailed 26-page specification with concrete protocol definition and mechanisms for implementation\",\n \"o\": 4,\n \"on\": \"Significant overlap with existing KEM-based Authentication for EDHOC draft, building directly upon that work\",\n \"mo\": 2,\n \"mon\": \"Single revision with no clear indication of working group adoption or broader community engagement\",\n \"r\": 1,\n \"rn\": \"This is a cryptographic protocol specification for device authentication, not related to AI or autonomous agents\",\n \"c\": [\"Other AI/agent\"]\n}", "in_tok": 649, "out_tok": 252} {"draft_name": "draft-cui-nmrg-auto-test", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft proposes an AI-assisted framework for automating network protocol testing using large language models, with components for protocol formalization, test generation, and iterative refinement. It defines 6 automation maturity levels (0-5) ranging from manual testing to fully autonomous adaptive systems.\",\n \"n\": 3,\n \"nn\": \"Applies established AI/LLM concepts to protocol testing domain with structured maturity levels - useful contribution but not groundbreaking\",\n \"m\": 2,\n \"mn\": \"Early framework sketch with high-level components and maturity definitions, but lacks detailed mechanisms, algorithms, or concrete implementation guidance\",\n \"o\": 2,\n \"on\": \"Some conceptual overlap with general AI testing frameworks but specific focus on network protocol testing appears relatively unique\",\n \"mo\": 2,\n \"mon\": \"Single draft submission with no indication of working group interest or community engagement beyond initial publication\",\n \"r\": 4,\n \"rn\": \"Directly addresses AI-assisted automation in networking domain with clear agent-like capabilities at higher maturity levels\",\n \"c\": [\"Autonomous netops\", \"Model serving/inference\"]\n}", "in_tok": 659, "out_tok": 271} {"draft_name": "draft-templin-manet-inet-omni", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft proposes a MANET internetworking framework using AERO and OMNI technologies to enable mobile ad-hoc networks to connect with fixed networks like the Internet. It builds on existing RFC 2501 MANET definitions to provide gateway and routing mechanisms.\",\"n\":2,\"nn\":\"Incremental work combining existing AERO/OMNI technologies for MANET internetworking, standard networking protocol development\",\"m\":3,\"mn\":\"Defines specific protocol framework and mechanisms for MANET internetworking with technical details\",\"o\":4,\"on\":\"Significant overlap with existing MANET routing protocols and internetworking approaches, builds heavily on established AERO/OMNI work\",\"mo\":2,\"mon\":\"Limited momentum indicated, appears to be individual submission with single revision\",\"r\":1,\"rn\":\"This is purely about mobile ad-hoc network routing and internetworking protocols, not related to AI or agents despite 'autonomous system' terminology\"}", "in_tok": 552, "out_tok": 214} {"draft_name": "draft-meunier-webbotauth-registry", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"Defines a JSON-based 'Signature Agent Card' format for web bots to advertise metadata about themselves including identity, purpose, rate limits, and cryptographic keys. Establishes an IANA registry for extensible discovery of agent information through a directory service.\",\n \"n\": 3,\n \"nn\": \"Useful contribution applying established agent card concepts specifically to web bot authentication with cryptographic signatures\",\n \"m\": 2,\n \"mn\": \"Early sketch with basic format description but lacking detailed protocol mechanics, implementation guidance, or concrete examples\",\n \"o\": 4,\n \"on\": \"Significant overlap with existing agent card formats and web authentication mechanisms, though specific combination may be novel\",\n \"mo\": 2,\n \"mon\": \"Single revision with future date suggests early stage work without demonstrated community engagement or working group adoption\",\n \"r\": 4,\n \"rn\": \"Directly relevant to AI agent authentication and discovery, addressing key challenges in bot identification and trust establishment\",\n \"c\": [\"Agent identity/auth\", \"Agent discovery/reg\", \"Data formats/interop\"]\n}", "in_tok": 558, "out_tok": 259} {"draft_name": "draft-chen-oauth-agent-revocation", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft extends OAuth 2.0 token revocation to support agent-based batch operations, cascade propagation, and audit trails for autonomous systems. It introduces new endpoints and protocols to handle revocation at the agent level rather than individual tokens.\",\n \"n\": 3,\n \"nn\": \"Useful contribution addressing practical gaps in OAuth for agent scenarios, though builds incrementally on existing revocation mechanisms\",\n \"m\": 3,\n \"mn\": \"Defines protocol mechanisms and new endpoints but appears to lack detailed implementation specifics and comprehensive examples\",\n \"o\": 2,\n \"on\": \"Minor conceptual overlap with general OAuth extensions but takes a distinct agent-centric approach to revocation\",\n \"mo\": 2,\n \"mon\": \"Single revision with future date suggests early stage work without demonstrated community engagement\",\n \"r\": 4,\n \"rn\": \"Directly addresses authentication and authorization needs for autonomous agent networks and cross-domain scenarios\",\n \"c\": [\"Agent identity/auth\", \"A2A protocols\"]\n}", "in_tok": 616, "out_tok": 243} {"draft_name": "draft-vandoulas-aidp", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"AIDP defines a control-plane protocol for secure agent interactions, delegation, and interoperability across heterogeneous systems. The protocol includes mechanisms for intent expression, authority enforcement, capability delegation, and auditable action execution.\",\"n\":4,\"nn\":\"Notable originality in creating a comprehensive control-plane protocol specifically for agent interactions with delegation and authority mechanisms\",\"m\":4,\"mn\":\"Detailed 34-page specification with defined protocol mechanisms, likely includes examples and substantial technical detail\",\"o\":2,\"on\":\"Minor similarities to existing protocols but unique focus on agent delegation and control-plane interactions\",\"mo\":2,\"mon\":\"Single revision with recent date suggests early stage without established community momentum\",\"r\":5,\"rn\":\"Core AI agent topic addressing fundamental needs for agent interoperability, delegation, and secure interactions\",\"c\":[\"A2A protocols\",\"Agent identity/auth\",\"Data formats/interop\",\"Policy/governance\"]}", "in_tok": 534, "out_tok": 199} {"draft_name": "draft-wang-jep-judgment-event-protocol", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"JEP defines a protocol for verifiable decision events with four immutable verbs (Judgment, Delegation, Termination, Verification) using signed JSON structures. It aims to provide interoperability for decision records across human, organizational, software, and autonomous agent systems.\",\n \"n\": 4,\n \"nn\": \"Novel approach to creating a universal protocol for verifiable decision events across diverse agent types with well-defined event semantics\",\n \"m\": 3,\n \"mn\": \"Defines core protocol structure with JSON/JWS but lacks detailed implementation examples and test vectors\",\n \"o\": 2,\n \"on\": \"Some conceptual overlap with event logging and audit trails, but unique focus on decision event verifiability across agent types\",\n \"mo\": 2,\n \"mon\": \"Single draft with future date suggests early stage development with no visible community engagement yet\",\n \"r\": 4,\n \"rn\": \"Directly addresses autonomous agent decision tracking and verification, key for multi-agent system accountability\",\n \"c\": [\"A2A protocols\", \"Data formats/interop\", \"Agent identity/auth\"]\n}", "in_tok": 678, "out_tok": 267} {"draft_name": "draft-kartha-grd", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft identifies the need for geospatial resource discovery services that enable applications like AR and autonomous systems to discover resources based on physical location rather than predefined identifiers. It presents a problem statement and conceptual architecture for spatial-based discovery mechanisms.\",\n \"n\": 4,\n \"nn\": \"Addresses a genuinely novel problem space where discovery begins with physical presence rather than known identifiers, which is fundamentally different from traditional name-based resolution\",\n \"m\": 1,\n \"mn\": \"Purely a problem statement and high-level conceptual architecture with no protocol definitions, data formats, or implementation details\",\n \"o\": 1,\n \"on\": \"Unique approach to spatial-based resource discovery that doesn't overlap significantly with existing internet protocols or standards\",\n \"mo\": 2,\n \"mon\": \"Recent draft from 2026 but appears to be initial work with no indication of working group adoption or broader community engagement yet\",\n \"r\": 4,\n \"rn\": \"Highly relevant to AI agents, particularly autonomous systems, robotics, and AR applications that need to discover resources based on spatial context rather than predetermined names\",\n \"c\": [\"Agent discovery/reg\", \"Autonomous netops\", \"Data formats/interop\"]\n}", "in_tok": 635, "out_tok": 292} {"draft_name": "draft-abbey-scim-agent-extension", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft extends SCIM 2.0 to support identity management for AI agents and agentic applications by adding new resource types and schemas. It aims to enable cross-domain agent identity management using existing SCIM infrastructure rather than creating new protocols.\",\n \"n\": 3,\n \"nn\": \"Useful contribution applying established identity management patterns to the emerging agent ecosystem\",\n \"m\": 2,\n \"mn\": \"Early sketch - abstract only provides high-level concepts without protocol details or concrete schemas\",\n \"o\": 2,\n \"on\": \"Minor similarities to general identity management approaches but unique focus on agent-specific SCIM extensions\",\n \"mo\": 1,\n \"mon\": \"Future-dated draft (2026) suggests this is conceptual work without current implementation momentum\",\n \"r\": 4,\n \"rn\": \"Directly addresses agent identity management needs, though focuses more on infrastructure than core AI capabilities\",\n \"c\": [\"Agent identity/auth\", \"Data formats/interop\"]\n}", "in_tok": 645, "out_tok": 241} {"draft_name": "draft-ietf-emu-hybrid-pqc-eapaka", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft enhances EAP-AKA' Forward Secrecy with hybrid post-quantum cryptography to protect against future quantum computer threats. It addresses the vulnerability of current ECDHE-based ephemeral key exchange by incorporating quantum-resistant algorithms.\",\n \"n\": 2,\n \"nn\": \"Incremental enhancement applying established post-quantum cryptography concepts to existing EAP-AKA' protocol\",\n \"m\": 3,\n \"mn\": \"Defined protocol enhancement with clear problem statement and solution approach, but may lack implementation details\",\n \"o\": 3,\n \"on\": \"Shares post-quantum cryptography concepts with other hybrid PQC drafts and builds on existing EAP-AKA' specifications\",\n \"mo\": 4,\n \"mon\": \"IETF working group draft with clear standardization path, addressing urgent quantum-readiness requirements\",\n \"r\": 1,\n \"rn\": \"Not about AI or agents - this is a network authentication protocol enhancement for quantum resistance\",\n \"c\": []\n}", "in_tok": 724, "out_tok": 255} {"draft_name": "draft-a2a-moqt-transport", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft specifies how the Agent-to-Agent (A2A) protocol can be transported over Media over QUIC Transport (MOQT), leveraging MOQT's publish/subscribe model and streaming capabilities for real-time AI agent communication. It addresses the transport layer requirements for A2A protocol operations including discovery, negotiation, and collaboration between AI agents.\",\"n\":2,\"nn\":\"Incremental work applying existing MOQT transport to A2A protocols - straightforward protocol mapping rather than novel approach\",\"m\":3,\"mn\":\"Appears to define specific protocol mechanisms for A2A over MOQT but limited detail available from abstract about implementation specifics\",\"o\":3,\"on\":\"Likely shares transport concepts with other QUIC-based drafts and A2A protocol work, though specific A2A-MOQT combination may be unique\",\"mo\":2,\"mon\":\"Single draft with October 2025 date suggests very early stage work with no clear community uptake yet\",\"r\":4,\"rn\":\"Directly addresses AI agent communication transport needs, which is core infrastructure for multi-agent systems\",\"c\":[\"A2A protocols\",\"Agent discovery/reg\",\"Data formats/interop\"]}", "in_tok": 561, "out_tok": 267} {"draft_name": "draft-ietf-httpbis-zstd-window-size", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft updates RFC8878 to make window size limits for Zstandard compression a requirement rather than recommendation in HTTP contexts. It addresses interoperability issues where browsers limit memory usage by restricting zstd window sizes.\",\n \"n\": 1,\n \"nn\": \"Trivial clarification converting existing recommendation to requirement with no new technical content\",\n \"m\": 4,\n \"mn\": \"Well-defined specification update with clear requirements and implementation guidance\",\n \"o\": 5,\n \"on\": \"Direct update to existing RFC8878 with complete overlap on the specific window sizing topic\",\n \"mo\": 4,\n \"mon\": \"IETF working group draft with clear adoption path addressing real interoperability problems\",\n \"r\": 1,\n \"rn\": \"Pure HTTP compression specification with no AI or agent-related content\",\n \"c\": [\"Other AI/agent\"]\n}", "in_tok": 541, "out_tok": 218} {"draft_name": "draft-zl-agents-networking-architecture", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft proposes a networking architecture for AI agents in enterprise and broadband environments, defining core components and their interactions. The document appears to be an early framework for how agents communicate and coordinate within network infrastructures.\",\"n\":3,\"nn\":\"Useful contribution addressing agent networking needs, though agent-based architectures are becoming common\",\"m\":2,\"mn\":\"Early sketch level - defines components and interactions but lacks detailed protocols or implementation specifics\",\"o\":3,\"on\":\"Shares networking architecture concepts with existing agent communication drafts but focuses on enterprise/broadband context\",\"mo\":2,\"mn\":\"Single revision with recent date, no evidence of working group adoption or broader community engagement yet\",\"r\":4,\"rn\":\"Directly relevant to AI agents with focus on networking architecture and agent interactions\",\"c\":[\"A2A protocols\",\"Autonomous netops\",\"Agent discovery/reg\"]}", "in_tok": 476, "out_tok": 188} {"draft_name": "draft-latour-pre-registration", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft proposes an EPP extension allowing TLD registries to perform pre-registration verification of domain names using AI/ML techniques to assign abuse risk scores from 0-100. The extension aims to proactively identify potentially abusive domain registrations before they are completed.\",\n \"n\": 3,\n \"nn\": \"Applies AI/ML scoring to existing EPP protocol for abuse detection - useful contribution combining established registry operations with modern ML approaches\",\n \"m\": 4,\n \"mn\": \"Detailed EPP protocol extension with XML schemas, command definitions, data flows and security considerations - appears implementation-ready\",\n \"o\": 2,\n \"on\": \"While domain abuse detection exists, the specific EPP pre-registration extension approach with ML scoring appears relatively unique\",\n \"mo\": 2,\n \"mon\": \"Single draft from May 2024 with no apparent working group adoption or broader community discussion yet\",\n \"r\": 4,\n \"rn\": \"Directly relevant as it specifies AI/ML integration into network protocols for automated decision-making in domain registration processes\",\n \"c\": [\"A2A protocols\", \"Policy/governance\", \"Autonomous netops\"]\n}", "in_tok": 664, "out_tok": 278} {"draft_name": "draft-yu-ai-agent-ipv6-networking-considerations", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft explores IPv6 networking considerations for AI agent communication, focusing on stable agent identifiers, forwarding policies, network context, and telemetry after discovery. It aims to complement existing agent protocols by addressing network-layer support rather than defining new discovery or collaboration mechanisms.\",\n \"n\": 3,\n \"nn\": \"Useful contribution applying IPv6 concepts to AI agent networking, though builds incrementally on established networking principles\",\n \"m\": 2,\n \"mn\": \"Early sketch stage with conceptual discussion but lacking detailed protocol specifications or implementation guidance\",\n \"o\": 2,\n \"on\": \"Minor overlap with general IPv6 and SRv6 work, but specific AI agent focus appears relatively unique\",\n \"mo\": 2,\n \"mon\": \"Single revision with future date suggesting early development phase, unclear community engagement\",\n \"r\": 4,\n \"rn\": \"Directly relevant to AI agent networking infrastructure, addressing important networking layer considerations for agent communication\",\n \"c\": [\"A2A protocols\", \"Agent identity/auth\", \"ML traffic mgmt\"]\n}", "in_tok": 723, "out_tok": 256} {"draft_name": "draft-wang-ppm-ecdh-psi", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft specifies a two-party Private Set Intersection protocol using Elliptic Curve Diffie-Hellman to enable participants to find common records without revealing non-intersecting private data. The protocol instantiates the Meadows match-making protocol with standard elliptic curves and hash-to-curve methods for privacy-preserving record matching.\",\n \"n\": 2,\n \"nn\": \"Incremental work applying well-known ECDH-based PSI techniques to IETF context - the underlying cryptographic approach is established\",\n \"m\": 4,\n \"mn\": \"Detailed technical specification with 31 pages covering protocol mechanics, though uncertain about test vectors and implementation details\",\n \"o\": 3,\n \"on\": \"PSI protocols exist in academic literature and other standards bodies, shares core concepts with established cryptographic PSI work\",\n \"mo\": 2,\n \"mon\": \"Single draft submission with future date (2025-08-20) suggests early stage, no clear working group adoption signals\",\n \"r\": 2,\n \"rn\": \"Privacy-preserving computation is tangentially related to AI/agent systems but this is primarily a general cryptographic protocol without specific AI focus\",\n \"c\": [\"Data formats/interop\"]\n}", "in_tok": 609, "out_tok": 302} {"draft_name": "draft-baysal-asimov-safety-architecture", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"Proposes a dual-gate safety architecture for autonomous AI agents with a deterministic denylist (Gate 1) and stateless LLM judge (Gate 2) to prevent self-manipulation of safety rules. The architecture separates reasoning from safety judgment through a four-layer priority hierarchy.\",\n \"n\": 4,\n \"nn\": \"Notable architectural insight of separating reasoning model from judging model to prevent conversational manipulation, with specific dual-gate implementation approach\",\n \"m\": 4,\n \"mn\": \"Detailed specification with components, layer semantics, conflict resolution rules, trust models, and conformance requirements across 27 pages\",\n \"o\": 2,\n \"on\": \"While AI safety is a common concern, the specific dual-gate architecture with stateless judge separation appears distinct from existing approaches\",\n \"mo\": 2,\n \"mon\": \"Single draft from 2026 with no indication of working group adoption or community feedback yet\",\n \"r\": 5,\n \"rn\": \"Directly addresses core AI agent safety concerns with action execution capabilities - fundamental to autonomous agent deployment\",\n \"c\": [\"AI safety/alignment\", \"Autonomous netops\", \"Policy/governance\"]\n}", "in_tok": 656, "out_tok": 283} {"draft_name": "draft-liu-rtgwg-llmsync-multicast", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft explores multicast networking applications for Large Language Model inference services and synchronization. The document appears to be an early examination of how multicast protocols could optimize LLM deployment architectures in cloud environments.\",\n \"n\": 3,\n \"nn\": \"Applying multicast to LLM inference represents a useful contribution addressing practical scaling challenges, though the core networking concepts are established\",\n \"m\": 2,\n \"mn\": \"At 6 pages focusing on use cases, this appears to be an early sketch without detailed protocol specifications or implementation guidance\",\n \"o\": 2,\n \"on\": \"While networking optimization for ML workloads exists, the specific focus on multicast for LLM synchronization has minor overlap with existing approaches\",\n \"mo\": 2,\n \"mon\": \"Single draft submission with future date suggests early stage work without demonstrated community engagement or working group adoption\",\n \"r\": 4,\n \"rn\": \"Directly addresses practical infrastructure concerns for LLM deployment and inference optimization, which is highly relevant to AI systems\",\n \"c\": [\"ML traffic mgmt\", \"Model serving/inference\"]\n}", "in_tok": 497, "out_tok": 264} {"draft_name": "draft-fu-cats-flow-lb", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft specifies a flow-level load balancing mechanism for Computing-Aware Traffic Steering (CATS) that reduces control plane overhead by enabling autonomous flow distribution in the data plane. The approach aims to improve resource utilization in computing-aware networks through distributed load balancing decisions.\",\"n\":2,\"nn\":\"Incremental extension of existing CATS framework with standard load balancing concepts applied to flow-level granularity\",\"m\":3,\"mn\":\"Defines specific protocol mechanism for flow-level load balancing with clear technical approach, though implementation details may need refinement\",\"o\":4,\"on\":\"Significant conceptual overlap with existing CATS drafts and general load balancing mechanisms, primarily differentiating on flow-level granularity\",\"mo\":3,\"mon\":\"Part of active CATS working group development with ongoing revisions and community engagement\",\"r\":4,\"rn\":\"Directly relevant to autonomous network operations and intelligent traffic management in computing-aware environments\",\"c\":[\"Autonomous netops\",\"ML traffic mgmt\"]}", "in_tok": 502, "out_tok": 223} {"draft_name": "draft-mozleywilliams-dnsop-bandaid", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"Proposes BANDAID, a DNS-based discovery mechanism for AI agents using existing DNS infrastructure and SVCB records. Defines structured namespace conventions and metadata exchange to enable agent-to-agent discovery without requiring DNS protocol changes.\",\"n\":3,\"nn\":\"Useful application of existing DNS mechanisms to emerging AI agent discovery needs, though the core concept of DNS-based service discovery is well-established\",\"m\":4,\"mn\":\"Detailed specification with clear namespace conventions, record formats, and operational procedures, appears implementation-ready\",\"o\":2,\"on\":\"While DNS service discovery exists, the specific application to AI agent discovery and the BANDAID framework appear relatively unique\",\"mo\":2,\"mon\":\"Single revision from October 2025 with no clear working group adoption or community discussion visible\",\"r\":4,\"rn\":\"Directly addresses AI agent infrastructure needs, specifically discovery and capability advertisement between autonomous agents\",\"c\":[\"Agent discovery/reg\",\"A2A protocols\",\"Agent identity/auth\"]}", "in_tok": 785, "out_tok": 214} {"draft_name": "draft-ahn-opsawg-5g-security-i2nsf-framework", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft proposes an integrated security framework for 5G networks that uses Intent-Based Networking (IBN) to translate high-level security intents into enforceable policies via the I2NSF architecture. The system supports adaptive policy enforcement across 5G core components and user equipment with closed-loop monitoring for autonomous security management.\",\n \"n\": 3,\n \"nn\": \"Useful integration of existing I2NSF and IBN concepts for 5G security, but combines known technologies rather than introducing fundamentally new approaches\",\n \"m\": 3,\n \"mn\": \"Presents a defined framework with clear architecture components and policy translation mechanisms, though lacks detailed protocol specifications and implementation examples\",\n \"o\": 3,\n \"on\": \"Shares concepts with existing I2NSF and IBN drafts, as well as 5G security frameworks, but focuses on their specific integration for edge security\",\n \"mo\": 2,\n \"mon\": \"Appears to be an early revision with limited evidence of working group adoption or broader community engagement in OPSAWG\",\n \"r\": 4,\n \"rn\": \"Directly relevant to autonomous network operations through IBN controllers that automatically translate intents to policies and provide closed-loop security management\",\n \"c\": [\"Autonomous netops\", \"Policy/governance\"]\n}", "in_tok": 667, "out_tok": 308} {"draft_name": "draft-han-agent-comm-enterprise", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft proposes technical mechanisms for AI agent communication in enterprise environments, including agent identification, registration, capability discovery, and semantic routing. It focuses on enabling trusted access and efficient collaboration among heterogeneous agents at campus scale.\",\"n\":2,\"nn\":\"Incremental application of existing networking concepts like service discovery and routing to AI agents in enterprise settings\",\"m\":2,\"mn\":\"Early sketch with proposed mechanisms but lacks detailed protocol specifications, message formats, or implementation guidance\",\"o\":4,\"on\":\"Significant conceptual overlap with existing service discovery, identity management, and enterprise networking drafts, just applied to AI agents\",\"mo\":2,\"mon\":\"Single revision draft with no evidence of working group adoption or broader community engagement\",\"r\":4,\"rn\":\"Directly addresses AI agent networking and communication challenges, though focused on enterprise deployment rather than broader internet scenarios\",\"c\":[\"A2A protocols\",\"Agent identity/auth\",\"Agent discovery/reg\"]}", "in_tok": 521, "out_tok": 201} {"draft_name": "draft-ietf-dnsop-ns-revalidation", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This IETF draft proposes an algorithm for DNS resolvers to revalidate NS record delegations by querying child zones directly and periodically refreshing delegation information. It aims to improve DNS resolution reliability by ensuring resolvers use the most authoritative and up-to-date nameserver information.\",\n \"n\": 2,\n \"nn\": \"Incremental improvement to existing DNS resolution practices, formalizing what some resolvers already do informally\",\n \"m\": 4,\n \"mn\": \"Detailed specification with clear algorithmic steps and implementation guidance for resolver behavior\",\n \"o\": 3,\n \"on\": \"Shares concepts with existing DNS operations drafts but addresses a specific gap in delegation handling\",\n \"mo\": 4,\n \"mn\": \"Active IETF working group draft with clear community interest in DNS operational improvements\",\n \"r\": 1,\n \"rn\": \"This is a DNS infrastructure specification with no relation to AI or autonomous agents\",\n \"c\": []\n}", "in_tok": 685, "out_tok": 238} {"draft_name": "draft-fv-rats-ear", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft defines EAT Attestation Result (EAR) message format for verifiers to encode appraisal results of attester evidence using AR4SI's trustworthiness vectors. EAR provides normalized evaluation results and contextual information to help relying parties make authorization decisions, supporting both simple and composite devices with CWT/JWT serialization.\",\n \"n\": 3,\n \"nn\": \"Useful contribution that extends existing EAT framework with standardized result format and trustworthiness vectors\",\n \"m\": 4,\n \"mn\": \"Detailed specification with clear message format definitions, examples, and support for both CWT and JWT serialization\",\n \"o\": 3,\n \"on\": \"Builds on existing EAT and AR4SI work, shares attestation concepts with related RATS WG drafts\",\n \"mo\": 4,\n \"mon\": \"Active development in RATS working group with clear progression toward standardization\",\n \"r\": 2,\n \"rn\": \"Generic attestation and trust framework not specifically designed for AI agents, though could potentially be applied to AI system verification\",\n \"c\": [\"Other AI/agent\"]\n}", "in_tok": 679, "out_tok": 277} {"draft_name": "draft-mcgraw-httpapi-agent-budget", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"Defines HTTP status code 427 (Budget Required) and Budget authentication scheme for software agents authorized to spend money or resources. Uses post-quantum signatures (ML-DSA-87) to provide cryptographic attestations of spending authorization separate from basic authentication.\",\n \"n\": 4,\n \"nn\": \"Novel separation of authentication vs budget authorization concerns with post-quantum cryptographic attestations for agent spending\",\n \"m\": 4,\n \"mn\": \"Detailed specification with concrete status codes, CBOR encoding, signature algorithms, and version negotiation mechanisms\",\n \"o\": 2,\n \"on\": \"Some conceptual similarity to HTTP payment protocols but unique approach combining agent authorization with post-quantum cryptography\",\n \"mo\": 2,\n \"mon\": \"Single draft from 2026 with no apparent working group adoption or community discussion\",\n \"r\": 5,\n \"rn\": \"Directly addresses core challenge of autonomous agent resource authorization and spending controls with practical HTTP integration\",\n \"c\": [\"Agent identity/auth\", \"A2A protocols\", \"Policy/governance\"]\n}", "in_tok": 694, "out_tok": 256} {"draft_name": "draft-ietf-idr-bgp-metrics", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft proposes extended per-hop BGP metrics to enable more sophisticated routing policies that are common in practice but currently require complex configurations. It supplements BGP-4 by adding globally significant metrics beyond the current limited set.\",\n \"n\": 2,\n \"nn\": \"Incremental improvement to BGP routing metrics - addresses practical operator needs but follows established patterns of BGP extensions\",\n \"m\": 3,\n \"mn\": \"Defines the protocol mechanism with clear problem statement and proposed solution, though may lack detailed implementation specifics\",\n \"o\": 3,\n \"on\": \"Likely shares concepts with other BGP extension drafts and traffic engineering proposals, but appears to have a specific focus on per-hop metrics\",\n \"mo\": 4,\n \"mon\": \"IETF IDR working group adoption suggests active development and community interest in BGP routing improvements\",\n \"r\": 1,\n \"rn\": \"This is a traditional BGP routing protocol extension with no apparent connection to AI, agents, or machine learning applications\",\n \"c\": [\"Other AI/agent\"]\n}", "in_tok": 588, "out_tok": 257} {"draft_name": "draft-patwhite-aauth", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft extends OAuth 2.1 with an Agent Authorization Grant for AI agents and software with persistent identities to obtain user consent and access tokens through polling, SSE, or WebSocket mechanisms. It builds on the OAuth Device Authorization Grant but adds agent-specific features like natural language descriptions and reasoning parameters.\",\"n\":3,\"nn\":\"Useful adaptation of existing OAuth patterns for agent use cases with meaningful extensions like natural language consent descriptions\",\"m\":2,\"mn\":\"Early sketch stage - only 6 pages with basic protocol outline but lacks detailed specification, examples, and security considerations\",\"o\":4,\"on\":\"Significant overlap with RFC 8628 Device Authorization Grant and general OAuth 2.1 framework, essentially a specialized variant\",\"mo\":2,\"mon\":\"Single draft from 2025 with no indication of working group adoption or community engagement\",\"r\":4,\"rn\":\"Directly addresses agent authentication needs - a core requirement for AI agent systems interacting with protected resources\",\"c\":[\"Agent identity/auth\",\"Human-agent interaction\"]}", "in_tok": 581, "out_tok": 224} {"draft_name": "draft-li-cats-intellinode-network-scheduling", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"IntelliNode introduces an in-network intelligent scheduling mechanism for CATS that uses FPGAs and programmable switches to perform real-time tensor-aware traffic steering for AI workloads. It creates a data-plane closed loop for perception-inference-decision-execution to optimize distributed AI training and inference through line-rate feature extraction and lightweight prediction models.\",\"n\":4,\"nn\":\"Notable integration of tensor semantics awareness into network data plane with FPGA-assisted real-time inference - goes beyond typical traffic steering\",\"m\":2,\"mn\":\"Early sketch with high-level architecture description but lacks detailed protocol specifications, message formats, or concrete implementation guidance\",\"o\":3,\"on\":\"Shares CATS framework foundation and some traffic management concepts with existing work, but tensor-aware data-plane approach is distinct\",\"mo\":2,\"mon\":\"Single revision draft from 2026 with no visible WG adoption or community discussion yet\",\"r\":4,\"rn\":\"Directly addresses AI workload optimization through intelligent network scheduling, highly relevant to AI agent infrastructure needs\",\"c\":[\"ML traffic mgmt\",\"Autonomous netops\"]}", "in_tok": 736, "out_tok": 242} {"draft_name": "draft-lin-opsec-trustroute-problem-statement", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft identifies requirements for differentiated security services in networks and explores trustworthiness-based routing as a solution. It focuses on extending current QoE/SLA guarantees to include security differentiation for various network scenarios.\",\n \"n\": 2,\n \"nn\": \"Incremental extension of existing QoS/SLA concepts to security domain, but trustworthiness-based routing adds some novelty\",\n \"m\": 1,\n \"mn\": \"Pure problem statement document with high-level discussion, no concrete protocols or mechanisms defined\",\n \"o\": 3,\n \"on\": \"Shares concepts with security-aware routing and QoS differentiation work, but specific trustworthiness angle is less common\",\n \"mo\": 2,\n \"mon\": \"Single draft from July 2024 with no visible follow-up or working group adoption\",\n \"r\": 2,\n \"rn\": \"Mentions trustworthiness concepts that could apply to AI systems, but primarily focused on traditional network security\",\n \"c\": [\"Other AI/agent\"]\n}", "in_tok": 591, "out_tok": 256} {"draft_name": "draft-ietf-lake-app-profiles", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft defines a canonical CBOR-based representation for describing and distributing EDHOC application profiles to enable interoperability. It provides coordination mechanisms for profile use and defines a set of well-known profiles for the lightweight authenticated key exchange protocol.\",\n \"n\": 2,\n \"nn\": \"Incremental work standardizing profile formats and coordination for existing EDHOC protocol\",\n \"m\": 4,\n \"mn\": \"Detailed specification with CBOR representations, coordination mechanisms, and concrete well-known profiles\",\n \"o\": 3,\n \"on\": \"Shares profile standardization concepts with other authentication and key exchange protocols\",\n \"mo\": 4,\n \"mon\": \"IETF working group draft with strong industry interest in lightweight authentication for IoT\",\n \"r\": 2,\n \"rn\": \"Authentication protocol standardization only tangentially related to AI agents through potential IoT device authentication\",\n \"c\": [\"Agent identity/auth\"]\n}", "in_tok": 689, "out_tok": 233} {"draft_name": "draft-kroehl-agentic-trust-aae", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"Defines Agent Authorization Envelope (AAE), a structured authorization container with MANDATE, CONSTRAINTS, and VALIDITY blocks for autonomous AI agents, using W3C DIDs and VCs. Addresses the gap where existing OAuth/API key mechanisms lack machine-evaluable semantics for agent actions without direct human oversight.\",\n \"n\": 4,\n \"nn\": \"Notable originality in defining machine-evaluable authorization semantics specifically for autonomous agents with structured mandate/constraint blocks\",\n \"m\": 4,\n \"mn\": \"Detailed specification with clear block structure, protocol bindings to W3C standards, and comprehensive coverage of authorization semantics\",\n \"o\": 2,\n \"on\": \"Minor similarities to general authorization frameworks but unique focus on autonomous agent-specific authorization containers\",\n \"mo\": 2,\n \"mon\": \"Single revision draft without evident working group adoption or community momentum signals\",\n \"r\": 5,\n \"rn\": \"Core AI agent topic directly addressing fundamental authorization challenges for production-scale autonomous agents\",\n \"c\": [\"Agent identity/auth\", \"Policy/governance\", \"Data formats/interop\"]\n}", "in_tok": 725, "out_tok": 266} {"draft_name": "draft-zhang-rtgwg-ai-agents-measurement", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft outlines use cases and requirements for communication protocols that enable measurement agents deployed on network devices to interact and coordinate. It focuses on the protocol needs for AI-driven measurement agents to collect, share, and act on network performance data.\",\"n\":2,\"nn\":\"Incremental work applying existing agent communication concepts to network measurement domain\",\"m\":2,\"mn\":\"Early stage focusing on use cases and requirements rather than concrete protocol specifications\",\"o\":3,\"on\":\"Shares concepts with existing network measurement and agent communication work but targets specific measurement agent use case\",\"mo\":2,\"mon\":\"Appears to be initial submission with limited evidence of broader community engagement\",\"r\":4,\"rn\":\"Directly relevant as it addresses AI agents in networking context with specific focus on measurement and monitoring\",\"c\":[\"A2A protocols\",\"Autonomous netops\",\"Data formats/interop\"]}", "in_tok": 477, "out_tok": 186} {"draft_name": "draft-scim-agent-extension", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft extends SCIM 2.0 to support identity management for AI agents and agentic applications. It proposes new resource types and schemas to enable interoperability between identity providers and agent systems.\",\n \"n\": 3,\n \"nn\": \"Applies existing identity management framework to agents - useful but incremental extension\",\n \"m\": 2,\n \"mn\": \"Early sketch with basic problem statement and approach, lacks detailed protocol specifications\",\n \"o\": 2,\n \"on\": \"Some conceptual overlap with general identity management but unique focus on agent identity\",\n \"mo\": 2,\n \"mon\": \"Future date (2026) suggests early/speculative work with limited current activity\",\n \"r\": 4,\n \"rn\": \"Directly addresses agent identity management, a key infrastructure need for AI systems\",\n \"c\": [\"Agent identity/auth\", \"Data formats/interop\"]\n}", "in_tok": 643, "out_tok": 221} {"draft_name": "draft-howard-rats-coserv", "model": "claude-sonnet-4-20250514", "raw": "```json\n{\n \"s\": \"CoSERV defines a structured query/result format for discovering and retrieving endorsements and reference values in remote attestation systems. It uses CDDL and CBOR serialization to enable efficient interoperability across diverse attestation infrastructure.\",\n \"n\": 3,\n \"nn\": \"Addresses a specific gap in RATS architecture with a structured approach, though query languages for resource discovery are well-established concepts\",\n \"m\": 4,\n \"mn\": \"Detailed specification with CDDL definitions and CBOR serialization format, appears implementation-ready\",\n \"o\": 2,\n \"on\": \"Builds on RATS architecture but appears to be the first to tackle this specific endorsement/reference value discovery problem\",\n \"mo\": 3,\n \"mon\": \"Active development with recent publication date, fits into established RATS working group ecosystem\",\n \"r\": 2,\n \"rn\": \"Remote attestation infrastructure that could support AI agents but not specifically focused on AI/agent systems\",\n \"c\": [\"Agent identity/auth\", \"Data formats/interop\"]\n}\n```", "in_tok": 583, "out_tok": 264} {"draft_name": "draft-ravikiran-clawdentity-protocol", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"Clawdentity defines a cryptographic identity and trust protocol specifically for AI agent-to-agent communication, featuring Ed25519 identities, registry-issued credentials, and authenticated relay transport. The protocol enables agents to prove identity and establish bilateral trust through pairing ceremonies without exposing private infrastructure.\",\n \"n\": 4,\n \"nn\": \"Notable originality in designing identity protocols specifically for AI agent communication patterns, addressing unique challenges of agent-to-agent trust establishment\",\n \"m\": 4,\n \"mn\": \"Detailed specification with specific cryptographic primitives, protocol flows, and transport mechanisms, appears implementation-ready\",\n \"o\": 2,\n \"on\": \"Minor similarities to general identity protocols but unique focus on AI agent communication scenarios with specialized pairing ceremonies\",\n \"mo\": 2,\n \"mon\": \"Single draft from 2026 with no indication of working group adoption or community momentum\",\n \"r\": 5,\n \"rn\": \"Core AI agent topic directly addressing fundamental identity and trust challenges in agent-to-agent communication\",\n \"c\": [\"Agent identity/auth\", \"A2A protocols\"]\n}", "in_tok": 586, "out_tok": 268} {"draft_name": "draft-dunbar-neotec-ac-pe2pe-ucmp-applicability", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft explores applying existing IETF YANG data models (AC, TE Topology, SR Policy, QoS) to support dynamic UCMP policy enforcement for AI inference traffic between distributed edge data centers. It serves as a practical evaluation of current IETF specifications for cloud-initiated, time-bounded network policy coordination.\",\"n\":2,\"nn\":\"Incremental application of existing YANG models to AI inference use case, not introducing new technical approaches\",\"m\":3,\"mn\":\"Defines specific application of existing models with clear use case but appears to be more analysis than implementation-ready specification\",\"o\":3,\"on\":\"Shares concepts with existing YANG model drafts and traffic engineering work, but specific AI inference application has some uniqueness\",\"mo\":2,\"mon\":\"Single revision with specific author, unclear WG involvement or broader community engagement\",\"r\":4,\"rn\":\"Directly addresses AI inference traffic management needs with specific focus on distributed AI workloads and dynamic network policy coordination\",\"c\":[\"ML traffic mgmt\",\"Autonomous netops\",\"Model serving/inference\"]}", "in_tok": 678, "out_tok": 235} {"draft_name": "draft-chen-agent-decoupled-authorization-model", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"Proposes a dynamic authorization framework for AI agents based on intent and behavioral trustworthiness rather than static identity/roles. Aims to provide just-in-time permissions that decouple authorization policies from business operations.\",\"n\":3,\"nn\":\"Intent-based authorization is established but applying it specifically to AI agent interactions with behavioral trust components represents a useful contribution\",\"m\":2,\"mn\":\"At 6 pages this appears to be an early framework sketch without detailed protocol mechanisms or implementation specifics\",\"o\":3,\"on\":\"Shares concepts with existing intent-based access control and zero-trust models, though AI agent focus provides some differentiation\",\"mo\":2,\"mon\":\"Single revision with recent date suggests early stage work without demonstrated community engagement\",\"r\":5,\"rn\":\"Directly addresses core AI agent authorization challenges, highly relevant to agent-to-agent interaction security\",\"c\":[\"Agent identity/auth\",\"A2A protocols\",\"AI safety/alignment\"]}", "in_tok": 535, "out_tok": 207} {"draft_name": "draft-ietf-rats-ar4si", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft defines standardized attestation result information elements that allow relying parties to evaluate the trustworthiness of attesters in heterogeneous environments. It provides CBOR and JSON serializations for consistent policy application across different attester and verifier types.\",\n \"n\": 3,\n \"nn\": \"Useful contribution to standardizing attestation results format, though builds on established RATS architecture concepts\",\n \"m\": 4,\n \"mn\": \"Detailed specification with concrete information model and two serialization formats, appears implementation-ready\",\n \"o\": 3,\n \"on\": \"Shares concepts with existing RATS framework drafts but focuses specifically on attestation result standardization\",\n \"mo\": 4,\n \"mon\": \"IETF working group draft with clear progression toward standardization in established RATS WG\",\n \"r\": 2,\n \"rn\": \"General attestation and trust framework - tangentially related to AI agents through trust establishment but not AI-specific\",\n \"c\": [\"Agent identity/auth\"]\n}", "in_tok": 576, "out_tok": 247} {"draft_name": "draft-cowles-volt", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"VOLT defines a tamper-evident logging protocol for AI agent operations using cryptographic hash chains and portable Evidence Bundles. It functions as a 'flight recorder' capturing agent workflow events with integrity guarantees while supporting privacy through content-addressed references and explicit redaction.\",\n \"n\": 4,\n \"nn\": \"Novel application of cryptographic audit trails specifically designed for AI agent workflows with privacy-first design and portable evidence bundles\",\n \"m\": 4,\n \"mn\": \"Detailed protocol specification at 40 pages with defined formats, hash chain mechanisms, and evidence bundle structure appears implementation-ready\",\n \"o\": 2,\n \"on\": \"While audit logging exists generally, the specific combination of hash chains, privacy preservation, and AI agent workflow tracing appears relatively unique\",\n \"mo\": 2,\n \"mon\": \"Single draft from 2026 with no clear indication of working group adoption or broader community engagement yet\",\n \"r\": 5,\n \"rn\": \"Directly addresses critical AI agent accountability and auditability needs with practical cryptographic integrity mechanisms for agent operations\",\n \"c\": [\"AI safety/alignment\", \"Data formats/interop\", \"Policy/governance\"]\n}", "in_tok": 672, "out_tok": 281} {"draft_name": "draft-ietf-grow-va", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft proposes Virtual Aggregation (VA) for FIB suppression to address routing table growth by selectively not loading certain RIB entries into the FIB. The approach allows ISPs to reduce FIB size by an order of magnitude while maintaining compatibility with existing infrastructure.\",\"n\":3,\"nn\":\"Useful contribution addressing practical FIB scalability challenges with a systematic suppression approach\",\"m\":4,\"mn\":\"Detailed specification with clear mechanisms, examples, and deployment considerations\",\"o\":2,\"on\":\"Some conceptual overlap with route aggregation techniques but distinct FIB-focused approach\",\"mo\":4,\"mon\":\"Active IETF GROW working group draft with clear industry problem focus\",\"r\":1,\"rn\":\"Traditional network routing optimization, not related to AI or autonomous agents\",\"c\":[\"Other AI/agent\"]}", "in_tok": 653, "out_tok": 186} {"draft_name": "draft-beyer-agent-identity-problem-statement", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft identifies the problem of establishing verifiable connections between software agents and the humans who authorize them, highlighting gaps in current identity systems for agent delegation and provenance. It frames the need for human-anchored agent identity without proposing specific solutions.\",\"n\":3,\"nn\":\"Useful contribution identifying an emerging problem space as agent ecosystems proliferate, though the core identity delegation concepts exist\",\"m\":1,\"mn\":\"Pure problem statement with no proposed mechanisms, protocols, or architectural solutions\",\"o\":2,\"on\":\"Some conceptual overlap with general identity delegation work but focused specifically on human-agent relationships\",\"mo\":2,\"mon\":\"Single revision from April 2026 with no indication of working group interest or follow-up activity\",\"r\":5,\"rn\":\"Directly addresses core AI agent infrastructure concerns around identity, delegation, and human oversight\",\"c\":[\"Agent identity/auth\",\"Human-agent interaction\",\"Policy/governance\"]}", "in_tok": 674, "out_tok": 199} {"draft_name": "draft-bondar-wca", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"Proposes a PKI-inspired infrastructure for cryptographically attesting the provenance of data sources consumed by AI agents through tool calls, addressing 'semantic laundering' where data acquires unwarranted trust from interface appearance. Defines Warrant Certificate Authorities (WCA) with graduated attestation levels and reference monitor properties for complete mediation of agent-tool interactions.\",\n \"n\": 4,\n \"nn\": \"Novel application of PKI trust models to AI agent provenance with original concept of 'semantic laundering' and graduated Warrant Attestation Levels\",\n \"m\": 4,\n \"mn\": \"Detailed specification with data structures, protocol definitions, trust hierarchy, and graduated adoption framework analogous to SLSA\",\n \"o\": 2,\n \"on\": \"Draws from established PKI and provenance concepts but applies them uniquely to AI agent context with minimal direct overlap\",\n \"mo\": 2,\n \"mn\": \"Future-dated draft (2026) suggests early stage with no apparent working group activity or community adoption yet\",\n \"r\": 5,\n \"rn\": \"Directly addresses core AI agent challenge of data provenance and trust in tool-call chains, highly relevant to agent security architecture\",\n \"c\": [\"Agent identity/auth\", \"AI safety/alignment\", \"Policy/governance\", \"Data formats/interop\"]\n}", "in_tok": 749, "out_tok": 319} {"draft_name": "draft-wu-savnet-inter-domain-architecture", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This document proposes an inter-domain SAVNET architecture for AS-level Source Address Validation (SAV) that enables autonomous systems to share SAV-specific information to generate more accurate validation rules. The architecture provides a framework for ASes to autonomously determine and exchange information needed for source address validation across domain boundaries.\",\"n\":3,\"nn\":\"Useful contribution to inter-domain source address validation with structured information sharing approach\",\"m\":2,\"mn\":\"Early architectural sketch with component definitions but lacks detailed protocol specifications\",\"o\":3,\"on\":\"Shares concepts with existing SAV and inter-domain routing security work but focuses specifically on information sharing architecture\",\"mo\":3,\"mon\":\"Active development as indicated by recent date, part of ongoing SAVNET working group efforts\",\"r\":3,\"rn\":\"Partially relevant as it involves autonomous decision-making by network systems for security validation, though not traditional AI agents\",\"c\":[\"Autonomous netops\"]}", "in_tok": 655, "out_tok": 204} {"draft_name": "draft-calabria-bmwg-ai-fabric-inference-bench", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft establishes standardized benchmarking methodologies for evaluating Ethernet-based AI inference serving network fabrics, focusing on disaggregated LLM architectures. It defines KPIs and test procedures for measuring fabric performance under realistic AI inference workloads including RDMA-based operations and MoE communication patterns.\",\"n\":3,\"nn\":\"Useful contribution applying established benchmarking principles to emerging AI inference fabric requirements, with some novel considerations for disaggregated architectures\",\"m\":4,\"mn\":\"Detailed specification with comprehensive test procedures, KPIs, and methodology covering multiple fabric scenarios though implementation examples could be enhanced\",\"o\":2,\"on\":\"Minor conceptual overlap with general network benchmarking standards but addresses unique AI inference fabric requirements not covered elsewhere\",\"mo\":3,\"mon\":\"Active development as evidenced by 2026 date and companion training draft, though community adoption level unclear from abstract\",\"r\":4,\"rn\":\"Directly relevant to AI infrastructure with focus on LLM inference performance measurement, though more infrastructure than agent-focused\",\"c\":[\"Model serving/inference\",\"ML traffic mgmt\"]}", "in_tok": 768, "out_tok": 245} {"draft_name": "draft-kuehlewind-audit-architecture", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"Proposes an architecture for auditing AI agent interactions on the Internet by linking intent, delegation, authorization, and execution through distributed audit records. The framework addresses gaps in existing auditing mechanisms that capture isolated events but miss delegation relationships and user intent in agent-driven systems.\",\n \"n\": 4,\n \"nn\": \"Novel approach to agent auditing that specifically addresses delegation chains and intent propagation, going beyond traditional system logging\",\n \"m\": 2,\n \"mn\": \"Early architectural sketch with basic components identified but lacking detailed protocol specifications or implementation details\",\n \"o\": 2,\n \"on\": \"Some conceptual overlap with general audit logging standards but unique focus on agent delegation and intent tracking\",\n \"mo\": 2,\n \"mn\": \"Single draft submission with no apparent working group adoption or community discussion yet\",\n \"r\": 5,\n \"rn\": \"Directly addresses core AI agent challenges around accountability, delegation, and auditable autonomous actions on the Internet\",\n \"c\": [\"Agent identity/auth\", \"Policy/governance\", \"A2A protocols\"]\n}", "in_tok": 590, "out_tok": 252} {"draft_name": "draft-ietf-lake-edhoc", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"EDHOC defines a compact authenticated Diffie-Hellman key exchange protocol optimized for constrained IoT environments. It provides mutual authentication, forward secrecy, and identity protection while reusing existing COSE, CBOR, and CoAP standards to minimize code footprint.\",\"n\":2,\"nn\":\"Incremental improvement combining existing cryptographic primitives and encoding standards for IoT constraints\",\"m\":5,\"mn\":\"Comprehensive 113-page specification with detailed protocol flows, message formats, cryptographic operations, and security considerations - implementation ready\",\"o\":3,\"on\":\"Shares key exchange concepts with TLS/DTLS and other authenticated DH protocols, but specific combination for constrained environments is distinctive\",\"mo\":5,\"mon\":\"IETF working group draft with strong industry backing for IoT security, addressing real deployment needs in constrained networks\",\"r\":2,\"rn\":\"Generic cryptographic protocol for IoT devices - only tangentially related to AI agents through potential use in securing agent communications\",\"c\":[\"Agent identity/auth\"]}", "in_tok": 598, "out_tok": 232} {"draft_name": "draft-haberkamp-ipp", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"IPP defines a cryptographic protocol for preserving verified human intent through chains of AI agent actions using signed tokens with lineage tracking. The protocol addresses accountability gaps in autonomous AI systems by creating tamper-evident audit trails from human principals to terminal agent actions.\",\"n\":4,\"nn\":\"Notable originality in applying cryptographic provenance specifically to human intent preservation in AI agent chains, with novel narrowing invariant concept\",\"m\":3,\"mn\":\"Well-defined protocol with clear properties and mechanisms, but lacks detailed implementation specifics and test vectors\",\"o\":2,\"on\":\"Minor conceptual overlap with general provenance and chain-of-custody protocols, but distinct focus on AI agent intent tracking\",\"mo\":2,\"mon\":\"Single revision draft with no apparent working group adoption or community feedback yet\",\"r\":5,\"rn\":\"Directly addresses core AI agent challenges around accountability, compliance, and human-AI interaction in enterprise environments\",\"c\":[\"Agent identity/auth\",\"Human-agent interaction\",\"Policy/governance\",\"A2A protocols\"]}", "in_tok": 770, "out_tok": 220} {"draft_name": "draft-helmprotocol-tttps", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"Proposes TTTPS protocol that extends TLS 1.3 with cryptographically verifiable temporal ordering to prevent manipulation by strategic channel controllers. Introduces Proof-of-Time mechanism specifically designed to address ordering attacks that are particularly acute for autonomous AI agents with symmetric reaction times.\",\n \"n\": 4,\n \"nn\": \"Novel approach combining temporal ordering proofs with TLS extension, addressing previously unformalized Strategic Channel Controller Problem with mathematical foundation\",\n \"m\": 3,\n \"mn\": \"Defined protocol with GRG integrity pipeline and AdaptiveSwitch mechanism, includes empirical validation but specification appears incomplete (truncated)\",\n \"o\": 2,\n \"on\": \"Some conceptual overlap with consensus protocols and timestamping services, but unique TLS integration and AI agent focus\",\n \"mo\": 3,\n \"mon\": \"Active development evidenced by deployment testing and empirical data collection, though experimental status indicates early adoption phase\",\n \"r\": 4,\n \"rn\": \"Directly addresses critical infrastructure problem for AI agent economies where ordering manipulation becomes economically strategic\",\n \"c\": [\"A2A protocols\", \"AI safety/alignment\", \"Autonomous netops\"]\n}", "in_tok": 932, "out_tok": 283} {"draft_name": "draft-mao-rtgwg-agent-comm-protocol-gap-analysis", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft analyzes gaps in existing networking and AI agent protocols for cross-device communication between AI agents deployed on network devices. It evaluates classical networking protocols and popular AI agent communication methods to identify what's missing for agent collaboration in network operations.\",\n \"n\": 2,\n \"nn\": \"Incremental gap analysis work that applies existing protocol evaluation methodology to the emerging domain of AI agent networking\",\n \"m\": 2,\n \"mn\": \"Early stage gap analysis document that identifies problems but doesn't propose specific solutions or detailed mechanisms\",\n \"o\": 3,\n \"on\": \"Shares concepts with related cross-device communication framework draft and general AI agent networking approaches\",\n \"mo\": 2,\n \"mon\": \"Single draft with future date suggests early development stage, unclear working group engagement or broader community interest\",\n \"r\": 4,\n \"rn\": \"Directly addresses AI agent communication protocols in network infrastructure, highly relevant to autonomous network operations\",\n \"c\": [\"A2A protocols\", \"Autonomous netops\", \"Data formats/interop\"]\n}", "in_tok": 616, "out_tok": 250} {"draft_name": "draft-zhao-nmop-network-management-agent", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft proposes Network Management Agents (NMA) as AI-driven entities that provide autonomous Level 4 network operations by adding cognitive capabilities like intent reasoning and planning on top of existing SDN controllers. It defines architectural integration modes and interface requirements for SDN-NMA interoperability.\",\n \"n\": 3,\n \"nn\": \"Useful contribution combining AI agents with SDN architecture, though builds incrementally on established autonomous network concepts\",\n \"m\": 3,\n \"mn\": \"Defines clear concepts and architecture with integration modes, but lacks detailed protocol specifications and implementation examples\",\n \"o\": 3,\n \"on\": \"Shares concepts with autonomous network management and AI-driven networking drafts, but offers distinct SDN integration approach\",\n \"mo\": 2,\n \"mon\": \"Early stage draft with single revision, no clear working group adoption signals yet\",\n \"r\": 5,\n \"rn\": \"Core AI agent topic directly addressing autonomous network operations with cognitive AI capabilities\",\n \"c\": [\"Autonomous netops\", \"A2A protocols\", \"Data formats/interop\"]\n}", "in_tok": 659, "out_tok": 260} {"draft_name": "draft-keller-aipref-vocab", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft proposes a standardized vocabulary for machine-readable opt-outs from AI training and text/data mining, defining three key use cases (TDM, AI Training, Generative AI Training) that can be referenced by opt-out systems. The vocabulary is mechanism-agnostic and aims to enable consistent interpretation of content usage restrictions across different implementations.\",\n \"n\": 2,\n \"nn\": \"Incremental work creating a simple taxonomy for existing opt-out concepts rather than introducing fundamentally new approaches\",\n \"m\": 2,\n \"mn\": \"Early sketch defining basic vocabulary terms but lacking detailed protocol mechanisms, implementation guidance, or concrete examples\",\n \"o\": 4,\n \"on\": \"Significant conceptual overlap with robots.txt, Creative Commons licensing, and other existing content usage restriction mechanisms\",\n \"mo\": 2,\n \"mon\": \"Single revision with no apparent working group adoption or broader community engagement\",\n \"r\": 4,\n \"rn\": \"Directly addresses AI training data governance, a core concern for AI systems and content creators seeking to control AI model training\",\n \"c\": [\"Policy/governance\", \"Data formats/interop\"]\n}", "in_tok": 578, "out_tok": 276} {"draft_name": "draft-chen-nmrg-semantic-inference-routing", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"SIRP proposes a semantic inference routing protocol that analyzes the content of AI inference requests to make intelligent routing decisions, rather than relying solely on client metadata. The framework includes value-added routing extensions for cost optimization, urgency handling, domain specialization, and privacy considerations.\",\n \"n\": 4,\n \"nn\": \"Content-level semantic analysis for AI inference routing represents notable originality beyond traditional network routing approaches\",\n \"m\": 3,\n \"mn\": \"Defines a complete protocol framework with semantic classification and routing mechanisms, though implementation details may need refinement\",\n \"o\": 2,\n \"on\": \"While AI inference serving exists, semantic content-based routing with VAR extensions shows minor similarities to existing work\",\n \"mo\": 3,\n \"mon\": \"Active development indicated by 2026 date, though community adoption and WG interest unclear from available information\",\n \"r\": 5,\n \"rn\": \"Directly addresses core AI agent infrastructure needs for intelligent request routing and inference optimization\",\n \"c\": [\"Model serving/inference\", \"ML traffic mgmt\", \"A2A protocols\"]\n}", "in_tok": 569, "out_tok": 264} {"draft_name": "draft-ietf-dnsop-dns-error-reporting", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft defines a lightweight DNS error reporting mechanism where validating resolvers automatically signal DNS resolution/validation failures to monitoring agents specified by authoritative servers. The error reporting uses extended DNS errors from RFC8914 and encodes the error information directly in the QNAME of the query itself.\",\"n\":2,\"nn\":\"Incremental improvement building on existing RFC8914 extended DNS errors, adding automated reporting capability\",\"m\":4,\"mn\":\"Well-defined protocol with clear mechanisms, message formats, and operational considerations though lacks detailed implementation examples\",\"o\":3,\"on\":\"Shares DNS operational monitoring concepts with other DNS management drafts but has distinct automated error reporting approach\",\"mo\":4,\"mon\":\"Active IETF dnsop working group draft with recent updates showing ongoing development interest\",\"r\":1,\"rn\":\"Pure DNS operational protocol for error reporting between DNS infrastructure components, not related to AI agents or autonomous systems\",\"c\":[\"Other AI/agent\"]}", "in_tok": 627, "out_tok": 206} {"draft_name": "draft-vandemeent-ains-discovery", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"AINS defines a protocol for discovering and establishing trust between autonomous agents (AI, devices, humans, services) using structured metadata records and cryptographic trust scores. Unlike DNS which maps names to addresses, AINS maps agent identifiers to rich capability and trust metadata objects with federated replication logs.\",\n \"n\": 4,\n \"nn\": \"Notable approach combining agent discovery with cryptographic trust resolution in a federated system, going beyond traditional naming services\",\n \"m\": 4,\n \"mn\": \"Detailed 34-page specification with structured record formats, HTTPS-based resolution protocol, and federated replication mechanism\",\n \"o\": 2,\n \"on\": \"While building on DNS concepts, the agent-focused metadata and trust resolution represents a distinct approach with minor similarities to existing naming systems\",\n \"mo\": 2,\n \"mon\": \"Single draft with future date (2026-03-29) and references to multiple companion protocols that may not exist yet\",\n \"r\": 5,\n \"rn\": \"Core AI agent infrastructure addressing fundamental needs for agent discovery, identification, and trust establishment in heterogeneous networks\",\n \"c\": [\"Agent discovery/reg\", \"Agent identity/auth\", \"A2A protocols\"]\n}", "in_tok": 671, "out_tok": 289} {"draft_name": "draft-mw-oauth-actor-chain", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft extends OAuth 2.0 Token Exchange with six profiles for preserving and cryptographically validating delegation chains across multi-hop service-to-service and agentic workflows. It addresses the gap where RFC 8693 allows nested actor claims but doesn't define how delegation paths are maintained and verified across successive token exchanges.\",\n \"n\": 4,\n \"nn\": \"Notable originality in defining comprehensive cryptographically verifiable delegation chain profiles for multi-hop workflows, addressing a real gap in OAuth 2.0 Token Exchange\",\n \"m\": 4,\n \"mn\": \"At 97 pages with six detailed profiles covering different disclosure and verification approaches, this appears to be a comprehensive specification with substantial technical detail\",\n \"o\": 2,\n \"on\": \"While building on OAuth 2.0 and Token Exchange foundations, the specific actor chain verification mechanisms and multi-hop delegation profiles appear relatively unique\",\n \"mo\": 2,\n \"mon\": \"Future-dated draft (2026-05-01) suggests this may be early-stage work, though the comprehensive nature indicates significant development effort\",\n \"r\": 4,\n \"rn\": \"Directly relevant to AI agent authentication and authorization, particularly for multi-hop agentic workflows where delegation path integrity is crucial for security and accountability\",\n \"c\": [\"Agent identity/auth\", \"A2A protocols\"]\n}", "in_tok": 733, "out_tok": 322} {"draft_name": "draft-yuan-rtgwg-traffic-agent-usecase", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This IETF draft proposes AI Network Traffic Optimization Agents as dynamic alternatives to static network optimization, covering use cases for tunnel adjustment, traffic steering, and network slice management. The document defines basic concepts and outlines three key application scenarios where autonomous AI agents can enhance network efficiency and reduce manual operations.\",\n \"n\": 2,\n \"nn\": \"Incremental application of well-known AI concepts to standard network optimization problems without novel technical approaches\",\n \"m\": 2,\n \"mn\": \"Early sketch that defines basic concepts and use cases but lacks detailed protocols, mechanisms, or implementation specifics\",\n \"o\": 4,\n \"on\": \"Significant conceptual overlap with existing network automation and AI-driven traffic management work in standards bodies\",\n \"mo\": 2,\n \"mon\": \"Single revision with future date suggests early-stage work without demonstrated community engagement or working group adoption\",\n \"r\": 4,\n \"rn\": \"Directly relevant to AI agents in networking context, focusing specifically on autonomous network traffic optimization scenarios\",\n \"c\": [\"ML traffic mgmt\", \"Autonomous netops\"]\n}", "in_tok": 656, "out_tok": 262} {"draft_name": "draft-zhu-anima-service-intent", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft defines Service Intent for Autonomic Networks, providing a structured semantic model to express service-level objectives and constraints across network, compute, and storage resources for emerging services like AI inference. It extends ANIMA Intent with service-oriented rather than connection-centric descriptions.\",\n \"n\": 3,\n \"nn\": \"Useful contribution extending ANIMA Intent to service-level objectives, particularly relevant for AI services, though builds incrementally on existing autonomic networking concepts\",\n \"m\": 3,\n \"mn\": \"Appears to define a concrete protocol mechanism with structured semantics, identification, and lifecycle management, though implementation details may need further elaboration\",\n \"o\": 3,\n \"on\": \"Shares concepts with existing ANIMA Intent work and general service orchestration approaches, but appears to have distinct focus on cross-resource service objectives\",\n \"mo\": 2,\n \"mon\": \"Single revision from 2026 with no indication of working group adoption or broader community engagement yet\",\n \"r\": 4,\n \"rn\": \"Directly relevant to AI agent operations by addressing service intent for AI inference workloads and autonomic network management of AI services\",\n \"c\": [\"Autonomous netops\", \"Model serving/inference\", \"Policy/governance\"]\n}", "in_tok": 557, "out_tok": 297} {"draft_name": "draft-chen-nmrg-multi-provider-inference-api", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft specifies extensions to the OpenAI API for multi-provider distributed AI inference, enabling provider diversity, load balancing, and failover while maintaining backward compatibility. The revision adds identity-based authorization, RBAC, and rate limiting for secure multi-tenant deployments.\",\n \"n\": 3,\n \"nn\": \"Useful contribution applying multi-provider orchestration patterns to AI inference APIs, though the concept of API federation and load balancing is well-established in other domains\",\n \"m\": 4,\n \"mn\": \"Detailed specification with concrete extensions, security mechanisms, and multi-tenant considerations spanning 49 pages, indicating implementation-ready detail\",\n \"o\": 3,\n \"on\": \"Shares architectural concepts with API gateway and service mesh approaches, and overlaps with other AI inference management proposals\",\n \"mo\": 2,\n \"mon\": \"Single revision (-01) with incremental updates, no clear indication of working group adoption or broader community engagement\",\n \"r\": 5,\n \"rn\": \"Core AI agent infrastructure topic addressing critical needs for provider diversity, scalability, and vendor independence in AI inference deployments\",\n \"c\": [\"Model serving/inference\", \"Agent identity/auth\", \"Data formats/interop\"]\n}", "in_tok": 678, "out_tok": 293} {"draft_name": "draft-luan-rtgwg-sdaf", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"Proposes Symmetry-Driven Asynchronous Forwarding (SDAF) for LEO satellite networks to address routing black holes and micro-loops by leveraging toroidal topology symmetry for microsecond-scale convergence without control-plane coordination. The mechanism is designed for resource-constrained satellites and claims compatibility with existing protocols like OSPF and IS-IS.\",\n \"n\": 4,\n \"nn\": \"Novel application of topology symmetry for autonomous forwarding decisions in satellite networks, representing a creative departure from traditional IGP-based approaches\",\n \"m\": 4,\n \"mn\": \"Well-defined mechanism with detailed protocol description, simulation results, and actual satellite router testing, though some implementation details could be more comprehensive\",\n \"o\": 2,\n \"on\": \"While fast reroute mechanisms exist, the symmetry-driven approach for satellite networks appears to have minimal overlap with existing IETF work\",\n \"mo\": 2,\n \"mon\": \"Single draft submission without clear working group adoption or broader community engagement visible\",\n \"r\": 3,\n \"rn\": \"Partially relevant as it involves autonomous network behavior and self-healing mechanisms, though focused on networking rather than AI agents specifically\",\n \"c\": [\"Autonomous netops\"]\n}", "in_tok": 886, "out_tok": 300} {"draft_name": "draft-liu-dmsc-acps-arc", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft proposes an Agent Collaboration Protocols (ACP) architecture for Internet of Agents, defining conceptual domains and functional components for agent interconnection. It covers agent registration, authentication, discovery, interaction, tool invocation, and monitoring to support large-scale agent collaboration.\",\"n\":3,\"nn\":\"Useful architectural framework for agent collaboration, though builds on established distributed systems and multi-agent concepts\",\"m\":2,\"mn\":\"Early architectural sketch with high-level components but lacks detailed protocol specifications, message formats, or concrete examples\",\"o\":3,\"on\":\"Shares conceptual overlap with existing multi-agent system architectures and distributed computing frameworks, but specific to IoT agent context\",\"mo\":2,\"mon\":\"Single revision draft with no indication of working group adoption or broader community engagement\",\"r\":5,\"rn\":\"Directly addresses core AI agent networking challenges including discovery, authentication, and collaboration protocols\",\"c\":[\"A2A protocols\",\"Agent identity/auth\",\"Agent discovery/reg\"]}", "in_tok": 609, "out_tok": 214} {"draft_name": "draft-nandakumar-ai-agent-moq-transport", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft proposes using MOQ Transport (MOQT) as a unified substrate for inter-agent communication protocols by providing an abstraction layer that maps request-response and streaming patterns to MOQT's publish/subscribe model. The approach is demonstrated with two protocols: Agent-to-Agent (A2A) for collaboration and Model Context Protocol (MCP) for tool access.\",\n \"n\": 3,\n \"nn\": \"Useful contribution applying existing MOQT to agent communication domain, though the core concept of protocol abstraction layers is well-established\",\n \"m\": 3,\n \"mn\": \"Defines the protocol abstraction and mapping mechanisms with concrete examples for A2A and MCP protocols, but lacks implementation details\",\n \"o\": 2,\n \"on\": \"Minor overlap with general transport protocol work, but unique focus on agent-specific communication patterns over MOQT\",\n \"mo\": 2,\n \"mon\": \"Single revision with future date suggests early stage, unclear community engagement or working group interest\",\n \"r\": 4,\n \"rn\": \"Directly addresses inter-agent communication infrastructure, a core requirement for AI agent ecosystems and multi-agent systems\",\n \"c\": [\"A2A protocols\", \"Data formats/interop\", \"Agent discovery/reg\"]\n}", "in_tok": 660, "out_tok": 305} {"draft_name": "draft-reilly-plpes", "model": "claude-sonnet-4-20250514", "raw": "```json\n{\n \"s\": \"PLPES defines a structured framework for standardizing AI prompts with machine-readable descriptors, classification taxonomy, versioning, and security requirements. It addresses the lack of interoperability standards and integrity mechanisms for prompts in enterprise and infrastructure AI deployments.\",\n \"n\": 4,\n \"nn\": \"Novel approach to treating prompts as protocol artifacts requiring formal specification - this is a genuinely new perspective on prompt engineering standardization\",\n \"m\": 4,\n \"mn\": \"Detailed specification with concrete PDO structure, five-tier taxonomy, versioning model, and security requirements - appears implementation-ready\",\n \"o\": 2,\n \"on\": \"Minor conceptual overlap with general AI safety practices, but the protocol-layer approach to prompt standardization is largely unique\",\n \"mo\": 1,\n \"mon\": \"Single draft from 2026 with no visible community engagement or working group adoption yet\",\n \"r\": 5,\n \"rn\": \"Directly addresses core AI agent infrastructure needs for prompt standardization, security, and interoperability in enterprise deployments\",\n \"c\": [\"Data formats/interop\", \"AI safety/alignment\", \"Policy/governance\"]\n}\n```", "in_tok": 775, "out_tok": 283} {"draft_name": "draft-levy-llm-uri-scheme", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft defines a URI scheme 'llm://' for standardizing connections to Large Language Model endpoints, encoding provider, model, credentials, and parameters in a single string similar to database connection URIs. The scheme aims to provide portable, standardized identification of LLM services across different providers and deployment scenarios.\",\n \"n\": 3,\n \"nn\": \"Useful standardization of LLM endpoint addressing, though conceptually straightforward adaptation of existing URI patterns to AI services\",\n \"m\": 4,\n \"mn\": \"Well-detailed specification with clear syntax definitions, parameter encoding, and practical examples for implementation\",\n \"o\": 2,\n \"on\": \"Novel application domain with minimal overlap, though builds on established URI scheme design principles\",\n \"mo\": 2,\n \"mon\": \"Single draft submission with no clear working group adoption or community discussion visible\",\n \"r\": 4,\n \"rn\": \"Directly addresses practical interoperability needs in AI agent and LLM integration scenarios\",\n \"c\": [\"Model serving/inference\", \"Data formats/interop\", \"Agent discovery/reg\"]\n}", "in_tok": 568, "out_tok": 262} {"draft_name": "draft-blahaj-grow-rpki-oauth", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft proposes using RPKI to authenticate parties in OpenID Connect flows for peering APIs, binding authentication to ASN operators. It addresses the lack of standardized authentication in the GROW Working Group's Peering API discussions.\",\"n\":3,\"nn\":\"Useful contribution combining RPKI trust anchoring with OIDC for network operator authentication - not obvious but builds on existing technologies\",\"m\":2,\"mn\":\"Early sketch stage with basic concept outlined but lacking detailed protocol mechanisms, implementation specifics, or comprehensive security analysis\",\"o\":2,\"on\":\"Minor conceptual similarities to general identity attestation work but unique application of RPKI to OIDC for ASN-based authentication\",\"mo\":3,\"mon\":\"Active development evidenced by recent date and GitHub repository, appears to be responding to ongoing GROW WG discussions\",\"r\":4,\"rn\":\"Directly relevant to agent identity and authentication in network automation contexts where ASN-bound agents need trusted identity attestation\",\"c\":[\"Agent identity/auth\",\"A2A protocols\"]}", "in_tok": 606, "out_tok": 227} {"draft_name": "draft-anandakrishnan-rats-ptv-agent-identity", "model": "claude-sonnet-4-20250514", "raw": "```json\n{\n \"s\": \"This draft proposes the PTV protocol for hardware-anchored attestation of AI agent identity, allowing agents to prove they're running authorized models without exposing sensitive data. The protocol defines CBOR message formats and composes with existing RATS attestation mechanisms for critical applications like healthcare and industrial control systems.\",\n \"n\": 4,\n \"nn\": \"Novel application of hardware attestation specifically to AI agent identity with separation of identity binding from behavioral continuity - addresses a gap in existing attestation frameworks\",\n \"m\": 3,\n \"mn\": \"Defines protocol structure, message types, and threat model but lacks detailed specification and implementation guidance for a complete solution\",\n \"o\": 2,\n \"on\": \"Builds on RATS attestation concepts but applies them specifically to AI agents in a novel way with minimal direct overlap\",\n \"mo\": 2,\n \"mon\": \"Recent draft date suggests early stage work without evidence of working group adoption or broader community engagement\",\n \"r\": 5,\n \"rn\": \"Directly addresses core AI agent security concerns around identity verification and model authorization - highly relevant to AI agent infrastructure\",\n \"c\": [\"Agent identity/auth\", \"AI safety/alignment\", \"Policy/governance\"]\n}\n```", "in_tok": 681, "out_tok": 290} {"draft_name": "draft-jurkovikj-collab-tunnel", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"TCT Protocol defines an HTTP-based method for efficient delivery of web content to automated agents using JSON envelopes that encapsulate resources in text formats like Markdown. It employs bidirectional URL discovery, JSON sitemaps, and conditional requests to optimize bandwidth while preserving semantic structure.\",\n \"n\": 3,\n \"nn\": \"Useful contribution combining existing HTTP mechanisms (ETags, conditional requests) with structured content delivery for agents, though individual components are not novel\",\n \"m\": 4,\n \"mn\": \"Detailed specification at 48 pages with defined protocol mechanisms, JSON schemas, and URL discovery patterns ready for implementation\",\n \"o\": 3,\n \"on\": \"Shares concepts with web content syndication formats and HTTP optimization techniques, but specific combination for agent content delivery is distinct\",\n \"mo\": 2,\n \"mon\": \"Single author draft without clear working group adoption or broader community engagement indicated\",\n \"r\": 4,\n \"rn\": \"Directly addresses automated agent content consumption with structured data formats and efficient transfer mechanisms\",\n \"c\": [\"Data formats/interop\", \"Human-agent interaction\", \"Agent discovery/reg\"]\n}", "in_tok": 573, "out_tok": 274} {"draft_name": "draft-ietf-v6ops-nd-considerations", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This document consolidates IPv6 Neighbor Discovery protocol issues from 20+ RFCs, identifying three root causes and proposing host isolation solutions. It provides guidance for selecting appropriate isolation methods to prevent ND security and operational problems in untrusted network environments.\",\"n\":2,\"nn\":\"Incremental consolidation work that organizes existing knowledge but doesn't introduce fundamentally new concepts\",\"m\":4,\"mn\":\"Well-developed analysis with detailed guidance and comprehensive coverage of existing solutions\",\"o\":3,\"on\":\"Builds upon extensive prior work in IPv6 ND but provides unique consolidation perspective\",\"mo\":4,\"mon\":\"Strong working group backing as IETF draft with clear operational need\",\"r\":1,\"rn\":\"Pure IPv6 networking protocol work with no AI or agent-related content\",\"c\":[\"Other\"]}", "in_tok": 689, "out_tok": 182} {"draft_name": "draft-cui-nmrg-llm-nm", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"Proposes a framework for integrating LLM agents into network management systems with mandatory human oversight and approval loops. Defines telemetry enhancements, agent decision modules, and interaction data models to enable collaborative automation while maintaining human control.\",\"n\":3,\"nn\":\"Useful contribution applying LLMs to network management with human-in-the-loop design, though the general concept of AI-assisted network operations with human oversight is becoming common\",\"m\":3,\"mn\":\"Defines framework components and workflows but appears to lack detailed protocol specifications, message formats, or concrete implementation guidance\",\"o\":2,\"on\":\"Some conceptual overlap with general AI-assisted network management approaches but the specific human-in-the-loop framework design appears relatively unique\",\"mo\":2,\"mn\":\"Single revision draft from April 2026 with no clear indication of working group adoption or broader community engagement\",\"r\":4,\"rn\":\"Directly addresses AI agents in network management with human oversight - highly relevant to autonomous network operations and human-agent interaction\",\"c\":[\"Autonomous netops\",\"Human-agent interaction\",\"Data formats/interop\"]}", "in_tok": 569, "out_tok": 241} {"draft_name": "draft-mpsb-agntcy-messaging", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft compares six messaging protocols for agentic AI systems, analyzing their streaming performance, delivery guarantees, security models, and operational complexity. It proposes AGNTCY SLIM as a purpose-built solution combining MLS with gRPC over HTTP/2 for quantum-safe end-to-end encryption and dynamic group membership for collaborative AI agents.\",\n \"n\": 3,\n \"nn\": \"Useful contribution applying existing messaging protocols to agentic AI context, though the core protocols and MLS integration are not novel\",\n \"m\": 2,\n \"mn\": \"Early sketch providing overview and comparison but lacking detailed protocol specification or implementation guidance for AGNTCY SLIM\",\n \"o\": 3,\n \"on\": \"Shares messaging infrastructure concepts with existing protocol drafts and MLS applications, but specific agentic AI focus is less common\",\n \"mo\": 2,\n \"mn\": \"Single revision from future date suggests early stage work without demonstrated community engagement or working group adoption\",\n \"r\": 4,\n \"rn\": \"Directly addresses core infrastructure needs for agentic AI systems including real-time collaboration, distributed coordination, and security requirements\",\n \"c\": [\"A2A protocols\", \"Agent identity/auth\"]\n}", "in_tok": 789, "out_tok": 293} {"draft_name": "draft-han-anima-ai-asa", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft proposes enhancing ANIMA's Autonomic Service Agents (ASA) with AI/LLM capabilities to improve intelligent network management functions. The document focuses on analyzing the communication protocol implications when ASAs become AI-powered.\",\n \"n\": 2,\n \"nn\": \"Incremental enhancement of existing ANIMA ASA framework with AI capabilities - straightforward application of LLMs to network automation\",\n \"m\": 2,\n \"mn\": \"Early conceptual stage with analysis of impacts but lacks concrete protocol specifications or implementation details\",\n \"o\": 3,\n \"on\": \"Builds directly on established ANIMA/GRASP protocols while sharing AI-enhanced network automation concepts with other autonomous networking approaches\",\n \"mo\": 2,\n \"mon\": \"Recently published single revision with unclear working group adoption status or broader community engagement\",\n \"r\": 4,\n \"rn\": \"Directly addresses AI-powered autonomous agents in networking context with focus on agent communication protocols\",\n \"c\": [\"Autonomous netops\", \"A2A protocols\", \"Agent discovery/reg\"]\n}", "in_tok": 575, "out_tok": 261} {"draft_name": "draft-fu-nmop-agent-communication-framework", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"Proposes a communication framework for multi-agent collaboration in network AIOps with AI gateway and Agent Name Service components. Aims to enable coordination of heterogeneous agents for intelligent network operations and maintenance.\",\"n\":3,\"nn\":\"Useful contribution applying multi-agent coordination concepts to network operations, though agent communication frameworks exist in other domains\",\"m\":2,\"mn\":\"Early sketch with high-level architecture description but lacks detailed protocol specifications, message formats, or concrete implementation details\",\"o\":3,\"on\":\"Shares concepts with existing agent discovery and communication work, though specific focus on network AIOps is somewhat distinct\",\"mo\":2,\"mon\":\"Single revision draft with no clear working group engagement or community feedback visible\",\"r\":5,\"rn\":\"Core AI agent topic directly addressing multi-agent coordination and communication for network operations\",\"c\":[\"A2A protocols\",\"Autonomous netops\",\"Agent discovery/reg\",\"Agent identity/auth\"]}", "in_tok": 602, "out_tok": 201} {"draft_name": "draft-aft-ai-traffic", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft identifies the problem of managing AI/LLM training and inference traffic across federated datacenters and edge locations. It focuses on the unique network characteristics and requirements of collective communication libraries used in distributed AI workloads.\",\n \"n\": 2,\n \"nn\": \"Problem is becoming well-recognized in industry; incremental contribution to document it formally\",\n \"m\": 1,\n \"mn\": \"Pure problem statement document with no proposed solutions or mechanisms\",\n \"o\": 3,\n \"on\": \"Shares concepts with general DC networking and ML infrastructure work, but specific AI traffic focus is somewhat distinct\",\n \"mo\": 3,\n \"mon\": \"Active area of interest given AI growth, but draft appears to be initial submission without established momentum\",\n \"r\": 4,\n \"rn\": \"Directly addresses AI/ML infrastructure challenges, though more focused on network transport than agent behavior\",\n \"c\": [\"ML traffic mgmt\", \"Model serving/inference\"]\n}", "in_tok": 606, "out_tok": 234} {"draft_name": "draft-embesozzi-oauth-agent-native-authorization", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft defines a structured metadata format for OAuth 2.0 First-Party Applications to enable interoperable authorization challenges for AI agents and non-browser clients. It extends the FiPA specification by adding an elicitations array to authorization challenge responses, using Model Context Protocol bindings for just-in-time agent authorization.\",\"n\":3,\"nn\":\"Useful contribution bridging OAuth 2.0 with AI agent authorization needs through structured metadata\",\"m\":3,\"mn\":\"Well-defined protocol extension with clear metadata format but lacks implementation examples\",\"o\":2,\"on\":\"Minor similarities to existing OAuth extensions but addresses unique AI agent authorization gap\",\"mo\":2,\"mon\":\"Single revision with 2026 date suggests early stage development\",\"r\":4,\"rn\":\"Directly addresses core AI agent authentication challenge in human-to-agent workflows\",\"c\":[\"Agent identity/auth\",\"Human-agent interaction\",\"Data formats/interop\"]}", "in_tok": 651, "out_tok": 203} {"draft_name": "draft-xu-sidrops-asrank-vulnerabilities", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft analyzes structural vulnerabilities in ASRank, an algorithm for inferring AS business relationships from BGP data, showing how small input changes can manipulate inference results in adversarial conditions. The document identifies the algorithm's sensitivity issues and discusses security implications with potential countermeasures.\",\n \"n\": 3,\n \"nn\": \"Useful contribution analyzing a specific vulnerability in widely-used BGP inference algorithms, though vulnerability analysis of existing systems is somewhat incremental\",\n \"m\": 3,\n \"mn\": \"Defines the vulnerability problem and provides analysis with examples, but appears to be more of an analysis document than a protocol specification\",\n \"o\": 2,\n \"on\": \"Specific focus on ASRank vulnerabilities appears relatively unique, with only minor conceptual overlap with general BGP security work\",\n \"mo\": 2,\n \"mon\": \"Single draft from 2026 with no clear indication of working group adoption or broader community engagement\",\n \"r\": 2,\n \"rn\": \"While analyzing algorithmic vulnerabilities, this is primarily about BGP routing inference rather than AI agents or autonomous systems in the AI context\",\n \"c\": [\"Policy/governance\", \"Other AI/agent\"]\n}", "in_tok": 595, "out_tok": 284} {"draft_name": "draft-jimenez-tbd-robotstxt-update", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft proposes updates to the robots.txt standard to better handle AI-specific crawlers by introducing new syntax for distinguishing between AI training and inference activities. It aims to give website owners more granular control over how AI systems access their content.\",\"n\":2,\"nn\":\"Incremental extension of existing robots.txt standard with AI-specific considerations\",\"m\":2,\"mn\":\"Early sketch - only 5 pages suggests limited detail on implementation specifics\",\"o\":4,\"on\":\"Significant overlap with existing robots.txt standards and various AI crawler management proposals\",\"mo\":2,\"mon\":\"Single revision with no evidence of working group adoption or community feedback\",\"r\":4,\"rn\":\"Directly relevant to AI agent governance and web crawling policy management\",\"c\":[\"Policy/governance\",\"Agent identity/auth\",\"Data formats/interop\"]}", "in_tok": 509, "out_tok": 183} {"draft_name": "draft-chen-lake-edhoc-aka", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft defines EDHOC-AKA, combining the EDHOC key exchange protocol with AKA authentication for efficient mobile network access authentication. It targets resource-constrained environments like Non-Terrestrial Networks using symmetric cryptography and pre-shared keys.\",\"n\":2,\"nn\":\"Incremental combination of two existing protocols (EDHOC and AKA) for mobile authentication scenarios\",\"m\":2,\"mn\":\"Early specification with basic protocol definition but lacks detailed implementation specifics and test vectors\",\"o\":4,\"on\":\"Significant overlap with existing EDHOC and AKA specifications, primarily combining established mechanisms\",\"mo\":2,\"mon\":\"Single revision draft with no clear working group adoption or community feedback visible\",\"r\":1,\"rn\":\"This is a mobile network authentication protocol, not related to AI agents or artificial intelligence systems\",\"c\":[\"Other AI/agent\"]}", "in_tok": 601, "out_tok": 195} {"draft_name": "draft-ietf-aipref-attach", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This IETF draft defines an HTTP mechanism for content creators to signal preferences about how automated systems (like AI) may consume their content. It updates RFC 9309 to include usage preference signaling during content acquisition.\",\"n\":3,\"nn\":\"Useful contribution addressing a timely need for content usage preferences in the AI era, though the HTTP signaling approach is relatively straightforward\",\"m\":4,\"mn\":\"Detailed specification that updates an existing RFC with clear mechanisms for preference signaling in HTTP\",\"o\":2,\"on\":\"Minor conceptual similarities to existing preference/policy mechanisms but addresses a distinct use case for AI content consumption\",\"mo\":5,\"mon\":\"Strong community momentum as an active IETF working group draft addressing urgent AI governance needs\",\"r\":4,\"rn\":\"Directly relevant to AI agent interactions with web content, though focuses more on content creator preferences than agent behavior\",\"c\":[\"Policy/governance\",\"Data formats/interop\",\"Human-agent interaction\"]}", "in_tok": 518, "out_tok": 211} {"draft_name": "draft-liu-agent-protocol-over-moq", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft specifies a framework for Agent-to-Agent communication over the Media over QUIC (MoQ) protocol, introducing application-layer framing for control signaling and session management. It aims to enable structured, low-latency communication between autonomous agents with support for both intra-domain and inter-domain deployment.\",\n \"n\": 3,\n \"nn\": \"Useful contribution applying existing MoQ transport to agent communication, but the core concept of agent-to-agent protocols over network transports is well-established\",\n \"m\": 2,\n \"mn\": \"Early sketch phase - abstract indicates framework design but lacks detail on specific protocol mechanisms, message formats, or implementation considerations\",\n \"o\": 3,\n \"on\": \"Shares conceptual space with other agent communication protocols and MoQ application drafts, though specific combination appears distinct\",\n \"mo\": 2,\n \"mon\": \"Future-dated draft (2026) suggests this is speculative work with no current implementation or community engagement\",\n \"r\": 4,\n \"rn\": \"Directly addresses core agent infrastructure needs for inter-agent communication, which is fundamental to multi-agent AI systems\",\n \"c\": [\"A2A protocols\", \"Data formats/interop\"]\n}", "in_tok": 567, "out_tok": 295} {"draft_name": "draft-yue-moq-transporting-sensing-data", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft proposes using Media Over QUIC (MOQ) to transport large-scale, real-time sensing data in 6G networks, addressing challenges in autonomous driving, smart cities, and industrial IoT. The approach focuses on leveraging QUIC's low-latency and multiplexing capabilities for efficient data fragmentation, secure transmission, and end-to-end QoS in next-generation wireless networks.\",\n \"n\": 2,\n \"nn\": \"Incremental application of existing MOQ protocol to sensing data transport; straightforward extension without novel protocol mechanisms\",\n \"m\": 2,\n \"mn\": \"Early sketch level - identifies challenges and proposes MOQ as solution but lacks detailed protocol specifications or concrete mechanisms\",\n \"o\": 3,\n \"on\": \"Shares transport optimization concepts with other networking drafts, though specific focus on sensing data is somewhat distinct\",\n \"mo\": 2,\n \"mn\": \"Single revision draft with no clear working group adoption or community engagement signals\",\n \"r\": 3,\n \"rn\": \"Partially relevant - enables AI/autonomous systems through sensing data transport but is primarily a networking transport solution rather than core AI agent technology\",\n \"c\": [\"ML traffic mgmt\", \"Data formats/interop\"]\n}", "in_tok": 680, "out_tok": 296} {"draft_name": "draft-tiloca-lake-edhoc-implem-cons", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This IETF draft provides implementation guidance for EDHOC, an authenticated key exchange protocol based on Ephemeral Diffie-Hellman Over COSE. The document offers practical considerations for developers implementing this cryptographic protocol in constrained environments.\",\"n\":2,\"nn\":\"Incremental contribution providing implementation guidance for existing EDHOC protocol\",\"m\":4,\"mn\":\"Detailed implementation considerations with specific guidance and examples\",\"o\":3,\"on\":\"Builds directly on base EDHOC specification with some overlap on protocol details\",\"mo\":3,\"mon\":\"Part of ongoing LAKE working group efforts with moderate activity\",\"r\":1,\"rn\":\"Pure cryptographic key exchange protocol, not related to AI or autonomous agents\",\"c\":[\"Other AI/agent\"]}", "in_tok": 506, "out_tok": 171} {"draft_name": "draft-feria-sas", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"Proposes a 'Reality Layer' framework to filter out synthetic noise from autonomous agents using deterministic admission control. Introduces the concept of Saturation Agentic Stridential (SAS) when agent-generated traffic overwhelms system capacity.\",\n \"n\": 4,\n \"nn\": \"Novel framing of agent noise pollution as a systemic problem requiring protocol-level solutions, with creative terminology and O(1) reality checking approach\",\n \"m\": 2,\n \"mn\": \"Early conceptual sketch with buzzwords like 'NIST-validated IRP-189' but lacks concrete protocol details, message formats, or implementation guidance\",\n \"o\": 2,\n \"on\": \"Unique problem formulation and solution approach, though shares general concerns about AI traffic management with other drafts\",\n \"mo\": 2,\n \"mon\": \"Single revision from 2026 with no indication of working group adoption or community feedback\",\n \"r\": 5,\n \"rn\": \"Directly addresses core challenge of autonomous agent systems overwhelming network infrastructure with synthetic traffic\",\n \"c\": [\"ML traffic mgmt\", \"AI safety/alignment\", \"Policy/governance\"]\n}", "in_tok": 599, "out_tok": 278} {"draft_name": "draft-zw-rtgwg-mcp-network-mgmt", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft extends the Model Context Protocol (MCP) JSON-RPC framework to enable network equipment to act as MCP servers for AI-driven network management. It defines capability tokens, tools, resources, and error codes to integrate network devices with AI controllers while maintaining backward compatibility.\",\"n\":3,\"nn\":\"Useful contribution applying existing MCP framework to network domain with appropriate extensions\",\"m\":3,\"mn\":\"Defines protocol extensions and mechanisms but lacks detailed implementation examples and test cases\",\"o\":2,\"on\":\"Minor conceptual overlap with network management protocols but unique application of MCP to networking\",\"mo\":2,\"mon\":\"Single revision with October 2025 date, no evidence of working group adoption or community engagement\",\"r\":4,\"rn\":\"Directly relevant to AI agents in network operations, enabling autonomous network management through standardized protocol\",\"c\":[\"Autonomous netops\",\"A2A protocols\",\"Data formats/interop\"]}", "in_tok": 553, "out_tok": 201} {"draft_name": "draft-hw-ai-agent-6g", "model": "claude-sonnet-4-20250514", "raw": "```json\n{\n \"s\": \"This draft analyzes requirements and enabling technologies for agent protocols specifically tailored to 6G mobile networks. It represents an early attempt to bridge AI agent communication protocols with telecommunications-specific performance and usage requirements.\",\n \"n\": 3,\n \"nn\": \"Useful contribution connecting agent protocols to 6G requirements, though the intersection of AI agents and telecom is becoming well-explored territory\",\n \"m\": 2,\n \"mn\": \"Early sketch phase - provides analysis and requirements but lacks concrete protocol definitions or detailed technical specifications\",\n \"o\": 3,\n \"on\": \"Shares concepts with other telecom AI integration work and general agent protocol discussions, but 6G-specific focus provides some differentiation\",\n \"mo\": 3,\n \"mon\": \"Active development with recent publication date, though momentum depends on 6G standardization timeline and telecom industry adoption\",\n \"r\": 4,\n \"rn\": \"Directly relevant as it focuses on agent protocols, though primarily from telecommunications infrastructure perspective rather than general agent-to-agent communication\",\n \"c\": [\"A2A protocols\", \"ML traffic mgmt\", \"Autonomous netops\", \"Policy/governance\"]\n}\n```", "in_tok": 582, "out_tok": 280} {"draft_name": "draft-vaughan-aipref-vocab", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft proposes a structured vocabulary for content rightsholders to express preferences about AI training use of their content through metadata and delivery protocols. The vocabulary covers permissions, usage scope, and data retention at various granularity levels to enable interoperability across Internet protocols.\",\"n\":2,\"nn\":\"Incremental extension of existing content preference mechanisms like robots.txt to the AI training domain\",\"m\":2,\"mn\":\"Early sketch stage with basic vocabulary outline but lacks detailed protocol specifications and implementation guidance\",\"o\":3,\"on\":\"Shares conceptual ground with web crawling standards and content licensing frameworks, but addresses AI-specific concerns\",\"mo\":2,\"mon\":\"Single revision draft without clear working group adoption or broad community engagement\",\"r\":4,\"rn\":\"Directly addresses the pressing need for AI training content governance and rightsholder control mechanisms\",\"c\":[\"Policy/governance\",\"Data formats/interop\"]}", "in_tok": 544, "out_tok": 195} {"draft_name": "draft-aip-agent-identity-protocol", "model": "claude-sonnet-4-20250514", "raw": "```json\n{\n \"s\": \"Defines Agent Identity Protocol (AIP) for verifiable AI agent authentication and policy enforcement through cryptographic identity and proxy-based access control. Addresses the problem of AI agents operating with unbounded permissions by creating identity boundaries between human and non-human actors.\",\n \"n\": 4,\n \"nn\": \"Notable originality in applying cryptographic identity specifically to AI agents with registry-based verification and policy enforcement proxy architecture\",\n \"m\": 4,\n \"mn\": \"Detailed specification with two-layer architecture, registry system, and enforcement mechanisms - appears implementation-ready\",\n \"o\": 2,\n \"on\": \"Minor similarities to general identity protocols but unique focus on AI agent-specific challenges and policy enforcement\",\n \"mo\": 2,\n \"mn\": \"Single revision dated 2026 suggests early stage development without established community momentum\",\n \"r\": 5,\n \"rn\": \"Core AI agent topic addressing fundamental authentication and authorization challenges for autonomous systems\",\n \"c\": [\"Agent identity/auth\", \"Policy/governance\", \"AI safety/alignment\"]\n}\n```", "in_tok": 649, "out_tok": 256} {"draft_name": "draft-mangrove-workgroup-mangrove", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"Mangrove proposes a unified framework for creating comprehensive Internet topology models by aggregating diverse data sources including traceroute, BGP, and IP allocation records. The system uses hierarchical abstraction and incremental updates to provide real-time Internet path estimation and performance metrics for research and operational analysis.\",\n \"n\": 3,\n \"nn\": \"Useful contribution combining existing measurement techniques into unified topology modeling, though individual components are well-established\",\n \"m\": 2,\n \"mn\": \"Early architectural sketch with high-level system description but lacks detailed protocol specifications or implementation details\",\n \"o\": 4,\n \"on\": \"Significant conceptual overlap with existing Internet measurement platforms like CAIDA, RIPE Atlas, and various topology mapping efforts\",\n \"mo\": 2,\n \"mn\": \"Single draft submission with no clear working group adoption or community feedback visible\",\n \"r\": 1,\n \"rn\": \"Traditional Internet measurement and topology mapping system with no apparent AI, ML, or autonomous agent components\",\n \"c\": [\"Other AI/agent\"]\n}", "in_tok": 773, "out_tok": 253} {"draft_name": "draft-kim-nmrg-2nmai5g", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft proposes using AI for autonomous security and network management in B5G networks through SOAR (Security Orchestration Automation and Response) technology. The document aims to define requirements for intent-based network management to ensure constant security quality.\",\"n\":2,\"nn\":\"Incremental application of existing AI concepts (SOAR, intent-based networking) to B5G without novel technical contributions\",\"m\":1,\"mn\":\"Abstract-level problem statement with poor English and unclear technical details, lacks concrete protocols or mechanisms\",\"o\":4,\"on\":\"Significant overlap with existing work on intent-based networking, AI-driven network management, and autonomous security systems\",\"mo\":1,\"mon\":\"Single draft from 2024 with no evidence of working group adoption or community interest\",\"r\":4,\"rn\":\"Directly relevant to AI agents in network operations and autonomous systems, though poorly articulated\",\"c\":[\"Autonomous netops\"]}", "in_tok": 594, "out_tok": 203} {"draft_name": "draft-aylward-aiga-2", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"AIGA specifies a comprehensive governance protocol for autonomous AI agents using tiered risk management, federated authority networks, and economic incentive alignment. The protocol includes immutable kernel architecture, action-based authorization, and advanced attestation mechanisms for high-risk scenarios.\",\n \"n\": 4,\n \"nn\": \"Novel integration of federated governance, immutable TCB architecture, and economic incentive alignment specifically for AI agent oversight - addresses unique challenges not covered by existing protocols\",\n \"m\": 4,\n \"mn\": \"Detailed specification with concrete mechanisms like M-TACE, warden triumvirate, and quarantine protocols - appears implementation-ready though lacks test vectors\",\n \"o\": 2,\n \"on\": \"While governance concepts exist elsewhere, the specific combination of AI agent oversight, federated authorities, and immutable kernel architecture is largely unique\",\n \"mo\": 2,\n \"mon\": \"Single revision (draft-02) with no clear WG adoption - appears to be individual work without demonstrated community engagement\",\n \"r\": 5,\n \"rn\": \"Directly addresses core AI agent governance challenges including autonomous operation, adversarial scenarios, and scalable oversight mechanisms\",\n \"c\": [\"Policy/governance\", \"Agent identity/auth\", \"AI safety/alignment\"]\n}", "in_tok": 792, "out_tok": 299} {"draft_name": "draft-zl-agents-networking-framework", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft introduces a networking framework for AI agents in enterprise and broadband environments, defining core components and their interactions. The document appears to be a high-level architectural framework rather than a detailed technical specification.\",\"n\":2,\"nn\":\"Incremental work applying existing networking concepts to agent environments without significant novel insights\",\"m\":2,\"mn\":\"Early sketch level with basic framework definition but lacking detailed protocols, mechanisms, or implementation guidance\",\"o\":3,\"on\":\"Shares general concepts with other agent networking and discovery drafts but appears to have distinct enterprise/broadband focus\",\"mo\":2,\"mon\":\"Single revision with recent publication date, no evidence of working group adoption or broader community engagement yet\",\"r\":4,\"rn\":\"Directly addresses AI agent networking infrastructure, which is highly relevant to current AI deployment challenges\",\"c\":[\"A2A protocols\",\"Autonomous netops\",\"Agent discovery/reg\"]}", "in_tok": 481, "out_tok": 192} {"draft_name": "draft-zhang-rtgwg-multicast-requirements-gaps-aidc", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft identifies requirements for multicast protocols in AI data centers, focusing on optimizing point-to-multipoint communications during large language model training and inference. It analyzes gaps between current multicast capabilities and the specific needs of AI workloads in data center environments.\",\n \"n\": 2,\n \"nn\": \"Incremental application of existing multicast concepts to AI data centers - useful but not particularly novel\",\n \"m\": 2,\n \"mn\": \"Early requirements and gap analysis document without detailed protocol mechanisms or specifications\",\n \"o\": 3,\n \"on\": \"Shares networking optimization concepts with other AI datacenter drafts but focuses specifically on multicast\",\n \"mo\": 2,\n \"mon\": \"Single draft submission with no clear working group adoption or broader community engagement\",\n \"r\": 3,\n \"rn\": \"Partially relevant - addresses networking infrastructure for AI but not core agent-to-agent communication protocols\",\n \"c\": [\"ML traffic mgmt\"]\n}", "in_tok": 537, "out_tok": 235} {"draft_name": "draft-zhang-dmsc-ioa-semantic-interaction", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft defines a semantic layer for agent-to-agent interaction in an Internet of Agents using ontology models and JSON-LD serialization. It specifies classes, properties, negotiation procedures, and alignment rules for semantic interoperability across heterogeneous agent domains.\",\"n\":3,\"nn\":\"Useful contribution applying established semantic web technologies to agent interaction, though the core concepts of ontology-based interoperability are well-established\",\"m\":3,\"mn\":\"Defines protocol mechanisms and procedures but lacks the detailed specifications and concrete examples needed for implementation\",\"o\":2,\"on\":\"While semantic interoperability has been explored, the specific application to Internet of Agents with JSON-LD profile appears relatively unique\",\"mo\":2,\"mn\":\"Appears to be early stage work without evidence of working group adoption or broader community engagement\",\"r\":5,\"rn\":\"Directly addresses core AI agent infrastructure needs for semantic interoperability and communication protocols\",\"c\":[\"A2A protocols\",\"Data formats/interop\",\"Agent discovery/reg\"]}", "in_tok": 600, "out_tok": 224} {"draft_name": "draft-scrm-aiproto-usecases", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft inventories use cases for Agentic AI systems that interact across networks, identifying protocol requirements for agent-to-agent communication, tool integration, and multi-domain coordination. It aims to guide IETF standardization efforts by analyzing concrete needs like long-lived interactions, delegation patterns, and security considerations.\",\n \"n\": 3,\n \"nn\": \"Useful contribution applying established use case methodology to emerging agentic AI domain, bridging AI capabilities with network protocol requirements\",\n \"m\": 2,\n \"mn\": \"Early-stage document focused on use case collection and requirement identification rather than concrete protocol specifications\",\n \"o\": 2,\n \"on\": \"References emerging A2A and MCP protocols but appears to be first comprehensive use case analysis specifically for IETF protocol implications\",\n \"mo\": 3,\n \"mon\": \"Active development with recent date, addresses timely topic of agentic AI protocols, though WG adoption status unclear\",\n \"r\": 5,\n \"rn\": \"Directly addresses core AI agent networking challenges including multi-agent coordination, tool interaction, and cross-domain operation\",\n \"c\": [\"A2A protocols\", \"Agent discovery/reg\", \"Data formats/interop\", \"Agent identity/auth\", \"Policy/governance\"]\n}", "in_tok": 697, "out_tok": 301} {"draft_name": "draft-yalcinkaya-rats-asil-m", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft defines an ASIL-M profile for RATS that enables authorization decisions based on evidence from multiple independent trust roots, specifically targeting AI inference systems. It introduces three key components: an Evidence Synthesis Protocol, the Twin Attestation Policy Language (TAPL), and Canonical Attestation Records (CAR) for audit and replay verification.\",\n \"n\": 4,\n \"nn\": \"Notable originality in addressing multi-root attestation synthesis for AI systems, combining evidence from independent trust anchors with deterministic policy evaluation\",\n \"m\": 3,\n \"mn\": \"Defines clear protocol mechanisms and artifacts but lacks detailed implementation specifics and test vectors for the three proposed components\",\n \"o\": 2,\n \"on\": \"Minor similarities to existing RATS work but the multi-root synthesis approach and AI-specific application profile are relatively unique\",\n \"mo\": 2,\n \"mn\": \"Single revision with future date suggests early stage development with uncertain community engagement\",\n \"r\": 4,\n \"rn\": \"Directly relevant to AI agent security and inference authorization, addressing critical attestation needs for AI system deployment\",\n \"c\": [\"Agent identity/auth\", \"AI safety/alignment\", \"Model serving/inference\"]\n}", "in_tok": 700, "out_tok": 288} {"draft_name": "draft-liao-aipref-autoctl-ext", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft extends an unspecified core protocol to provide granular server-side controls for automated interactions, including rate limiting, technology restrictions, and permission management. It aims to give service owners better control over how AI agents and automation systems interact with their services.\",\n \"n\": 2,\n \"nn\": \"Incremental extension focusing on server-side automation controls - useful but not particularly original given existing rate limiting and permission frameworks\",\n \"m\": 2,\n \"mn\": \"Early sketch stage - references undefined [CORE-SPEC], lacks concrete protocol details, mechanisms, or implementation guidance\",\n \"o\": 4,\n \"on\": \"Significant overlap with existing rate limiting protocols, API permission frameworks, and web automation control mechanisms like robots.txt\",\n \"mo\": 1,\n \"mon\": \"Appears inactive with undefined core specification reference and minimal development indicators\",\n \"r\": 4,\n \"rn\": \"Directly relevant to AI agent interactions - addresses practical needs for controlling automated systems accessing web services\",\n \"c\": [\"Policy/governance\", \"Agent identity/auth\", \"ML traffic mgmt\"]\n}", "in_tok": 518, "out_tok": 261} {"draft_name": "draft-stone-atxn", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"Defines a cryptographically-signed transaction protocol for agent-to-agent value exchanges that maps to existing payment infrastructure while providing legal defensibility under current contract law. Part of a comprehensive multi-draft framework for agentic commerce including escrow, dispute resolution, and agent identity verification.\",\"n\":4,\"nn\":\"Novel approach to bridging AI agent commerce with existing legal and financial infrastructure through canonical transaction bundles\",\"m\":4,\"mn\":\"Detailed 27-page specification with concrete rail mappings, conformance tiers, and integration with companion protocols\",\"o\":2,\"on\":\"Unique focus on legally-defensible agent transactions, minimal overlap with existing payment or identity protocols\",\"mo\":3,\"mon\":\"Active development with coordinated multi-draft submission suggesting sustained effort, but unclear WG adoption status\",\"r\":5,\"rn\":\"Directly addresses core challenge of autonomous agent commerce with practical legal and technical framework\",\"c\":[\"A2A protocols\",\"Agent identity/auth\",\"Policy/governance\"]}", "in_tok": 861, "out_tok": 215} {"draft_name": "draft-williams-intent-token", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"Specifies Intent Token, a cryptographic authorization primitive that binds autonomous AI agent actions to human-declared authorization envelopes before execution. Addresses gaps in existing OAuth/OIDC frameworks which handle session-level access but not action-level authorization for autonomous agents.\",\n \"n\": 4,\n \"nn\": \"Notable originality in bridging human intent declaration with autonomous agent execution through cryptographic binding - addresses a genuine gap between existing session-based auth and action-level governance\",\n \"m\": 2,\n \"mn\": \"Early sketch level - presents the concept and problem statement but likely lacks detailed protocol mechanics, message formats, and implementation specifics needed for a complete specification\",\n \"o\": 2,\n \"on\": \"Minor similarities to existing authorization frameworks but the specific focus on cryptographically binding human intent to autonomous agent actions appears relatively unique\",\n \"mo\": 2,\n \"mon\": \"Single revision from 2026 draft with no clear indication of working group adoption or broader community engagement yet\",\n \"r\": 5,\n \"rn\": \"Core AI agent topic directly addressing fundamental authorization challenges for autonomous systems - highly relevant to AI agent infrastructure\",\n \"c\": [\"Agent identity/auth\", \"AI safety/alignment\", \"Policy/governance\"]\n}", "in_tok": 604, "out_tok": 293} {"draft_name": "draft-mcp-over-moqt", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft specifies how to transport Model Context Protocol (MCP) messages over Media over QUIC Transport (MOQT) to enable efficient communication between AI language models and external data sources. It maps MCP's request-response and streaming patterns onto MOQT's publish-subscribe object delivery model over QUIC/WebTransport.\",\"n\":2,\"nn\":\"Incremental work combining two existing protocols - straightforward transport layer adaptation\",\"m\":4,\"mn\":\"Detailed 32-page specification with message mapping procedures and session management\",\"o\":3,\"on\":\"Shares transport concepts with other QUIC-based protocols but specific MCP/MOQT combination is distinct\",\"mo\":2,\"mon\":\"Single revision with specific October 2025 date, unclear adoption status\",\"r\":4,\"rn\":\"Directly relevant to AI agent infrastructure - addresses core transport needs for model-external system integration\",\"c\":[\"A2A protocols\",\"Data formats/interop\",\"Model serving/inference\"]}", "in_tok": 573, "out_tok": 220} {"draft_name": "draft-yl-agent-id-requirements", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"Proposes requirements for digital identity management in AI agent communication protocols, focusing on cross-industry interoperability and security. Aims to establish foundational identity mechanisms for agent-to-agent and agent-to-entity communication.\",\n \"n\": 2,\n \"nn\": \"Requirements document for applying existing digital identity concepts to AI agents - incremental adaptation rather than novel approach\",\n \"m\": 2,\n \"mn\": \"Early requirements sketch at 9 pages - lacks detailed protocol specifications, implementation guidance, or concrete technical mechanisms\",\n \"o\": 4,\n \"on\": \"Significant overlap with existing digital identity frameworks (OAuth, DID, PKI) and agent communication standards - mostly adaptation work\",\n \"mo\": 2,\n \"mon\": \"Single revision draft with no indication of working group adoption or broader community engagement in AI agent identity space\",\n \"r\": 4,\n \"rn\": \"Directly addresses core challenge in AI agent ecosystems - identity management is fundamental for secure multi-agent communication\",\n \"c\": [\"Agent identity/auth\", \"A2A protocols\"]\n}", "in_tok": 562, "out_tok": 257} {"draft_name": "draft-ietf-lamps-est-renewal-info", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft extends RFC7030 EST to provide server-controlled recommendations for when certificates should be renewed, replacing the typical 50% lifetime rule. It addresses the need for more flexible renewal timing as certificate lifetimes decrease from years to months or shorter periods.\",\"n\":2,\"nn\":\"Incremental extension to existing EST protocol, addressing a practical timing issue but not introducing novel concepts\",\"m\":3,\"mn\":\"Appears to define a specific protocol mechanism with clear problem statement, though implementation details unclear from abstract\",\"o\":3,\"on\":\"Likely shares renewal timing concepts with other certificate management protocols and PKI standards\",\"mo\":4,\"mon\":\"IETF working group adoption suggests active development and community interest in certificate lifecycle management\",\"r\":1,\"rn\":\"This is a certificate management protocol extension with no relation to AI or agents - appears to be a false positive for AI/agent relevance\",\"c\":[\"Other AI/agent\"]}", "in_tok": 573, "out_tok": 202} {"draft_name": "draft-ietf-ace-edhoc-oscore-profile", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This IETF draft specifies an ACE framework profile that uses EDHOC for mutual authentication between OAuth clients and resource servers, establishing OSCORE security contexts for protecting communications. It enables delegation of authorization management from resource-constrained servers to more capable trusted hosts.\",\"n\":2,\"nn\":\"Incremental combination of existing protocols (EDHOC, OSCORE, ACE) for constrained environments\",\"m\":4,\"mn\":\"Detailed 89-page specification with comprehensive protocol definitions and examples\",\"o\":3,\"on\":\"Builds on established ACE, EDHOC, and OSCORE standards but creates new integration profile\",\"mo\":4,\"mon\":\"Active IETF working group draft with 2026 target date indicating strong development momentum\",\"r\":2,\"rn\":\"Focuses on IoT device authentication and authorization rather than AI agents specifically\",\"c\":[\"Agent identity/auth\"]}", "in_tok": 686, "out_tok": 201} {"draft_name": "draft-kamimura-rats-behavioral-evidence", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft explores the conceptual relationship between remote attestation (verifying system trustworthiness) and behavioral evidence recording (tracking what systems actually do), discussing how these complementary approaches could work together in accountability frameworks. It's purely informational and doesn't propose new protocols or modifications to existing RATS architecture.\",\"n\":2,\"nn\":\"Incremental conceptual discussion connecting existing attestation concepts with behavioral monitoring - useful but not groundbreaking\",\"m\":2,\"mn\":\"Early conceptual sketch that explicitly avoids defining mechanisms, protocols, or cryptographic bindings\",\"o\":3,\"on\":\"Builds directly on RFC 9334 RATS architecture and shares behavioral monitoring concepts with some security frameworks\",\"mo\":2,\"mon\":\"Single revision draft with no clear working group adoption or broader community engagement yet\",\"r\":2,\"rn\":\"Only tangentially related to AI/agents - focuses on general system attestation and behavioral evidence without specific AI agent considerations\",\"c\":[\"Policy/governance\"]}", "in_tok": 614, "out_tok": 211} {"draft_name": "draft-yang-dynamic-ecn-threshold", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft proposes dynamically adjusting ECN marking thresholds to align with dynamic buffer allocation in network devices. It provides a deterministic algorithm to maintain adaptive relationships between congestion marking and actual buffer capacity without requiring machine learning models.\",\"n\":3,\"nn\":\"Useful contribution addressing real misalignment problem between static thresholds and dynamic buffers, though the concept of adaptive thresholds isn't entirely new\",\"m\":4,\"mn\":\"Detailed specification with operational framework, deterministic reference algorithm, and configurable parameters - appears implementation-ready\",\"o\":2,\"on\":\"Some conceptual overlap with general AQM and buffer management work, but specific approach to ECN-dynamic buffer coupling is relatively unique\",\"mo\":2,\"mn\":\"Single revision from 2026 with no indication of working group adoption or broader community engagement yet\",\"r\":2,\"rn\":\"Tangentially AI-related only through mention of avoiding ML models - primarily a network buffer management mechanism rather than AI/agent technology\",\"c\":[\"Other AI/agent\"]}", "in_tok": 673, "out_tok": 224} {"draft_name": "draft-ietf-lake-edhoc-psk", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft specifies a Pre-Shared Key authentication method for the EDHOC key exchange protocol, enabling efficient mutual authentication and session resumption with quantum resistance. The PSK method reduces computational overhead while maintaining security properties including identity protection and ephemeral key exchange.\",\"n\":2,\"nn\":\"Incremental extension adding PSK authentication to existing EDHOC protocol - useful but straightforward adaptation\",\"m\":4,\"mn\":\"Detailed specification with complete protocol flows, key derivation procedures, message formats, and comprehensive security considerations\",\"o\":3,\"on\":\"Shares core EDHOC concepts with base protocol draft, but PSK authentication method has distinct characteristics\",\"mo\":4,\"mon\":\"IETF working group draft with clear adoption path and active development trajectory\",\"r\":2,\"rn\":\"General cryptographic protocol for secure communications - not specifically designed for AI agents despite potential applicability\",\"c\":[\"Other AI/agent\"]}", "in_tok": 624, "out_tok": 202} {"draft_name": "draft-agentir-aepp", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"AEPP defines a standardized protocol for executing AI agents and handling payments on the internet, providing a missing execution layer for existing agent URI schemes. It implements a three-phase model (discovery, payment, execution) with on-chain verification and claims to serve 1.8M agents with zero per-agent fees.\",\n \"n\": 4,\n \"nn\": \"Notable originality in combining agent execution standardization with integrated payment verification and bridging multiple existing agent URI schemes\",\n \"m\": 4,\n \"mn\": \"Detailed specification with live reference implementation serving significant scale, though lacks formal test vectors\",\n \"o\": 2,\n \"on\": \"Minor similarities to existing payment and execution protocols but unique in agent-specific integration approach\",\n \"mo\": 4,\n \"mn\": \"Strong momentum evidenced by live implementation at scale and integration with multiple existing agent naming schemes\",\n \"r\": 5,\n \"rn\": \"Core AI agent topic addressing fundamental infrastructure needs for agent execution and economic coordination\",\n \"c\": [\"A2A protocols\", \"Agent discovery/reg\", \"Model serving/inference\"]\n}", "in_tok": 600, "out_tok": 262} {"draft_name": "draft-gaikwad-llm-benchmarking-profiles", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft defines concrete benchmarking profiles for LLM serving systems, specifying System Under Test boundaries, measurement points, and interpretation constraints. It builds on companion drafts for terminology and methodology to enable reproducible performance comparisons.\",\n \"n\": 3,\n \"nn\": \"Useful contribution addressing the practical need for standardized LLM benchmarking approaches, though the concept of performance profiles is well-established in other domains\",\n \"m\": 4,\n \"mn\": \"Detailed specification with clear profile definitions, measurement boundaries, and concrete examples at 53 pages, appears implementation-ready\",\n \"o\": 3,\n \"on\": \"Shares conceptual framework with general benchmarking standards and references companion LLM benchmarking drafts, but applies established concepts to LLM-specific context\",\n \"mo\": 3,\n \"mon\": \"Part of active multi-draft series on LLM benchmarking with recent publication date, showing ongoing development effort\",\n \"r\": 4,\n \"rn\": \"Directly addresses critical need for standardized LLM performance evaluation, essential for deployment and comparison of AI systems\",\n \"c\": [\"Model serving/inference\"]\n}", "in_tok": 602, "out_tok": 280} {"draft_name": "draft-toorop-dnsop-ranking-dns-data", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This document extends RFC2181's DNS data trustworthiness ranking by adding entries for root server data and DNSSEC-verified data, while assigning numerical values to ranking positions. It aims to provide a more complete and quantifiable framework for comparing DNS data trustworthiness.\",\"n\":2,\"nn\":\"Incremental extension of existing RFC2181 ranking system with logical additions\",\"m\":2,\"mn\":\"Early specification with basic framework but lacks detailed implementation guidance\",\"o\":4,\"on\":\"Directly builds upon and extends RFC2181's existing trustworthiness ranking system\",\"mo\":2,\"mon\":\"Single revision draft with no clear working group momentum or community engagement\",\"r\":1,\"rn\":\"Standard DNS protocol enhancement unrelated to AI or autonomous agents\",\"c\":[\"Other\"]}", "in_tok": 581, "out_tok": 173} {"draft_name": "draft-teodor-pilot-protocol", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"Pilot Protocol creates an overlay network providing autonomous AI agents with virtual addressing, transport services, NAT traversal, and encrypted communication tunnels over UDP. The protocol aims to give AI agents first-class network citizenship with stable identities and reachable addresses, operating beneath application protocols like A2A and MCP.\",\n \"n\": 4,\n \"nn\": \"Creating a dedicated overlay network specifically for AI agent communication with virtual addressing and bilateral trust models represents notable originality in agent networking\",\n \"m\": 4,\n \"mn\": \"At 43 pages with detailed protocol specification including addressing, transport, NAT traversal, and encryption mechanisms, this appears to be a comprehensive and detailed specification\",\n \"o\": 2,\n \"on\": \"While overlay networks exist, the specific focus on AI agent communication needs and integration with emerging agent protocols like A2A/MCP shows minor similarities to existing work\",\n \"mo\": 2,\n \"mon\": \"As a single draft from 2026 with no indication of working group adoption or broader community engagement, momentum appears limited to initial publication\",\n \"r\": 5,\n \"rn\": \"Directly addresses core AI agent networking infrastructure needs including identity, addressing, transport, and secure communication - central to agent-to-agent interaction\",\n \"c\": [\"A2A protocols\", \"Agent identity/auth\", \"Agent discovery/reg\"]\n}", "in_tok": 595, "out_tok": 319} {"draft_name": "draft-larsson-aitlp", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"Defines AITLP protocol for organizational governance of autonomous software agents, focusing on identity, trust, lifecycle management and authority boundaries rather than capabilities. Complements existing agent communication protocols by providing organizational actor framework with cryptographic scope enforcement.\",\"n\":4,\"nn\":\"Notable focus on organizational boundaries and 'what agents cannot do' rather than capabilities, with novel hierarchical mandate enforcement and Agent Legacy Mode for knowledge transfer\",\"m\":3,\"mn\":\"Well-defined protocol with clear identity conventions, lifecycle states, and trust verification mechanisms, though implementation details and concrete examples could be stronger\",\"o\":2,\"on\":\"Explicitly positions itself as complementary to MCP, A2A, and ANS rather than overlapping, addressing orthogonal governance concerns\",\"mo\":2,\"mon\":\"Recent draft date suggests early stage, unclear if part of working group or has broader community engagement\",\"r\":5,\"rn\":\"Directly addresses core challenge of autonomous agent deployment in organizational contexts with focus on auditability and revocation\",\"c\":[\"Agent identity/auth\",\"Policy/governance\",\"A2A protocols\"]}", "in_tok": 661, "out_tok": 234} {"draft_name": "draft-zhao-detnet-enhanced-use-cases", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft extends RFC8578 by defining new use cases for deterministic networks including industrial internet, high-experience video, intelligent computing, and ISAC-enabled smart factories. It outlines network requirements and common properties for scaling deterministic networks beyond the original DetNet scope.\",\"n\":2,\"nn\":\"Incremental extension of existing DetNet work with new application domains\",\"m\":2,\"mn\":\"Early stage problem statement and use case description without detailed technical solutions\",\"o\":4,\"on\":\"Significant overlap with existing DetNet RFCs and related deterministic networking work\",\"mo\":3,\"mon\":\"Active development as indicated by 2026 date, likely part of ongoing DetNet evolution\",\"r\":3,\"rn\":\"Partially relevant through intelligent computing and smart factory automation use cases that may involve AI agents\",\"c\":[\"ML traffic mgmt\",\"Autonomous netops\"]}", "in_tok": 510, "out_tok": 190} {"draft_name": "draft-liu-oauth-a2a-profile", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft defines an OAuth profile for Agent-to-Agent communication that embeds call chain context within Transaction Tokens to maintain agent identity and authorization across distributed agent workloads. It extends existing OAuth Transaction Token mechanisms specifically for A2A scenarios in trusted domains.\",\"n\":2,\"nn\":\"Incremental extension of existing OAuth Transaction Tokens to A2A context - useful but straightforward application\",\"m\":2,\"mn\":\"Early specification stage at only 5 pages - likely contains basic profile definition but lacks detailed implementation guidance\",\"o\":3,\"on\":\"Builds directly on OAuth Transaction Tokens RFC work and shares concepts with other agent authentication approaches\",\"mo\":2,\"mon\":\"Single draft submission with no clear working group adoption signals yet\",\"r\":4,\"rn\":\"Directly addresses agent-to-agent authentication and authorization, which is fundamental for distributed AI agent systems\",\"c\":[\"Agent identity/auth\",\"A2A protocols\"]}", "in_tok": 526, "out_tok": 200} {"draft_name": "draft-drake-email-tpm-attestation", "model": "claude-sonnet-4-20250514", "raw": "```json\n{\n \"s\": \"Proposes hardware TPM attestation in email headers to verify senders have genuine security chips, making large-scale spam economically infeasible by requiring unique physical hardware per sender. Includes both direct TPM verification and privacy-preserving SD-JWT alternatives for Sybil-resistant authentication.\",\n \"n\": 4,\n \"nn\": \"Novel application of TPM attestation to email authentication with specific focus on AI-generated spam prevention - creative bridging of hardware security and messaging protocols\",\n \"m\": 4,\n \"mn\": \"Detailed 25-page specification with defined mechanisms, header formats, and verification procedures, though lacks implementation examples and test vectors\",\n \"o\": 2,\n \"on\": \"Limited overlap - builds on existing TPM attestation and SD-JWT standards but applies them uniquely to email authentication problem\",\n \"mo\": 2,\n \"mn\": \"Single revision draft with future date suggests early stage, no clear working group adoption or community feedback yet\",\n \"r\": 4,\n \"rn\": \"Directly addresses AI-generated content challenges by making automated spam campaigns require substantial hardware investment, highly relevant to AI agent authentication\",\n \"c\": [\"Agent identity/auth\", \"Policy/governance\"]\n}\n```", "in_tok": 731, "out_tok": 293} {"draft_name": "draft-yao-agent-auth-considerations", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft proposes extensions to OAuth for authenticating and authorizing AI agents within Agent Communication Networks (ACN). It addresses the critical need for scalable and trustworthy authentication mechanisms as AI agents become more prevalent in various industries.\",\"n\":3,\"nn\":\"Useful contribution applying established OAuth framework to emerging AI agent authentication needs\",\"m\":2,\"mn\":\"Early sketch stage - proposes workflows but lacks detailed protocol specifications and implementation details\",\"o\":3,\"on\":\"Shares authentication concepts with existing OAuth and identity management drafts but focuses on AI agent-specific use cases\",\"mo\":2,\"mon\":\"Single revision with future date suggests early development stage, no clear working group adoption\",\"r\":5,\"rn\":\"Core AI agent topic directly addressing fundamental authentication and authorization requirements for agent networks\",\"c\":[\"Agent identity/auth\",\"A2A protocols\",\"Policy/governance\"]}", "in_tok": 521, "out_tok": 186} {"draft_name": "draft-allman-tcp-early-rexmt", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"Proposes Early Retransmit mechanism for TCP and SCTP to recover lost segments when congestion window is small by reducing duplicate ACKs needed for fast retransmission. Aims to avoid lengthy retransmission timeouts in low-bandwidth or small-window scenarios.\",\"n\":2,\"nn\":\"Incremental improvement to existing TCP fast retransmit mechanism, addressing edge case of small congestion windows\",\"m\":4,\"mn\":\"Detailed specification with clear mechanism definition, specific conditions for triggering, and implementation guidance\",\"o\":3,\"on\":\"Shares concepts with existing TCP congestion control and retransmission mechanisms but proposes specific enhancement\",\"mo\":1,\"mon\":\"This appears to be Mark Allman's work from early 2000s, likely inactive given the author and timeframe\",\"r\":1,\"rn\":\"Pure TCP transport protocol optimization with no relation to AI, agents, or machine learning systems\",\"c\":[\"Other AI/agent\"]}", "in_tok": 559, "out_tok": 214} {"draft_name": "draft-sovereign-satp", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"SATP proposes a protocol framework for establishing verifiable identity, attribution, and governance for autonomous machines including AI agents. It aims to provide a non-repudiable root of trust for both digital AI agents and physical autonomous systems.\",\"n\":4,\"nn\":\"Notable approach to autonomous machine identity and governance, addressing an emerging need as AI agents become more prevalent\",\"m\":2,\"mn\":\"Early sketch level - only abstract provided, lacks protocol details, mechanisms, or implementation specifics\",\"o\":2,\"on\":\"Some conceptual overlap with existing identity and trust frameworks, but specific focus on autonomous machines is relatively unique\",\"mo\":2,\"mon\":\"Single revision with recent date, but no evidence of working group adoption or broader community engagement\",\"r\":5,\"rn\":\"Core AI agent topic directly addressing fundamental challenges of agent identity, trust, and governance in autonomous systems\",\"c\":[\"Agent identity/auth\",\"Policy/governance\"]}", "in_tok": 517, "out_tok": 198} {"draft_name": "draft-jewell-aibdp", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"AIBDP defines a declarative protocol for websites to specify AI usage boundaries, extending robots.txt-style mechanisms to control how AI systems can index, train on, and interact with web content. The protocol aims to provide machine-readable permissions and restrictions for various AI activities including training, mimicry, and agentic access.\",\"n\":3,\"nn\":\"Useful contribution extending existing web crawling control mechanisms to AI-specific use cases, though the core concept of declarative content usage policies is not entirely novel\",\"m\":3,\"mn\":\"Appears to define specific protocol mechanisms and likely includes examples, though without seeing full technical details it's positioned as a defined protocol rather than just conceptual\",\"o\":3,\"on\":\"Shares concepts with existing robots.txt standards and likely overlaps with other web content control mechanisms, but applies them to the specific AI domain\",\"mo\":2,\"mon\":\"Single revision from 2026 date suggests early stage development without demonstrated working group adoption or broader community momentum yet\",\"r\":4,\"rn\":\"Directly addresses a core AI governance challenge of how content creators can control AI system access to their materials, highly relevant to current AI deployment concerns\",\"c\":[\"Policy/governance\",\"Data formats/interop\"]}", "in_tok": 602, "out_tok": 266} {"draft_name": "draft-miller-ztip", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"ZTIP defines primitives for verifiable delegation, intent binding, and behavioral attestation in multi-agent systems using zero-trust principles. The 49-page specification appears to provide a comprehensive protocol for secure agent interactions and verification.\",\"n\":4,\"nn\":\"Zero-trust applied to intent-based multi-agent systems is a notably original approach combining emerging concepts\",\"m\":4,\"mn\":\"49 pages suggests detailed specification with substantial protocol definition and likely examples\",\"o\":2,\"on\":\"While touching on zero-trust and agent concepts, the specific intent protocol approach appears relatively unique\",\"mo\":2,\"mon\":\"Single revision suggests early stage without demonstrated community engagement\",\"r\":5,\"rn\":\"Multi-agent systems with verifiable delegation and behavioral attestation is core to AI agent infrastructure\",\"c\":[\"A2A protocols\",\"Agent identity/auth\",\"AI safety/alignment\"]}", "in_tok": 482, "out_tok": 189} {"draft_name": "draft-ye-problems-and-requirements-of-dns-for-ioa", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft analyzes problems with DNS in supporting Internet of Agents (IoA) scenarios and explores technical requirements for agent collaboration. It focuses on how DNS needs to evolve to accommodate AI-driven agent interactions and discovery.\",\n \"n\": 3,\n \"nn\": \"Useful contribution applying DNS to agent systems, though DNS evolution for new paradigms is an established pattern\",\n \"m\": 2,\n \"mn\": \"Early sketch focusing on problem statement and requirements analysis without detailed technical solutions\",\n \"o\": 2,\n \"on\": \"Minor similarities to general DNS evolution work but unique focus on agent-specific requirements\",\n \"mo\": 2,\n \"mon\": \"Single draft with future date suggests early stage work without demonstrated community engagement\",\n \"r\": 4,\n \"rn\": \"Directly relevant to AI agent infrastructure, addressing core networking needs for agent discovery and collaboration\",\n \"c\": [\"Agent discovery/reg\", \"A2A protocols\", \"Data formats/interop\"]\n}", "in_tok": 517, "out_tok": 235} {"draft_name": "draft-ietf-lamps-rfc7030-csrattrs", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft clarifies ambiguities in RFC7030's CSR Attributes Response specification for EST (Enrollment over Secure Transport), addressing implementation challenges through clearer encoding rules. It introduces a template-based approach for CSR contents that allows EST servers to specify subject Distinguished Names and other attributes more precisely.\",\n \"n\": 2,\n \"nn\": \"Incremental clarification of existing protocol ambiguities with straightforward template approach\",\n \"m\": 4,\n \"mn\": \"Detailed specification with clear encoding rules and practical template mechanism for implementation\",\n \"o\": 4,\n \"on\": \"Significant overlap with RFC7030 and RFC9148 as this is an update/clarification document\",\n \"mo\": 4,\n \"mon\": \"IETF working group draft with recent publication date suggesting active development\",\n \"r\": 1,\n \"rn\": \"Pure PKI certificate enrollment protocol specification with no AI or agent-related content\",\n \"c\": [\"Other\"]\n}", "in_tok": 691, "out_tok": 239} {"draft_name": "draft-illyes-rep-purpose", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft extends the Robots Exclusion Protocol (robots.txt) to allow targeting of web crawlers and automated clients based on their stated purpose for accessing content, rather than just their user-agent string. It proposes a new rule type that could enable more granular control over different types of AI agents and automated systems based on their intended use case.\",\"n\":3,\"nn\":\"Useful extension addressing real need as AI agents diversify beyond traditional web crawlers\",\"m\":2,\"mn\":\"Early specification stage with basic concept outlined but lacking detailed implementation guidance\",\"o\":2,\"on\":\"Builds directly on RFC9309 but addresses distinct targeting mechanism not covered elsewhere\",\"mo\":2,\"mon\":\"Single revision draft with no clear working group activity or community discussion\",\"r\":4,\"rn\":\"Directly relevant to AI agent ecosystem management and web resource access control\",\"c\":[\"ML traffic mgmt\",\"Policy/governance\",\"Agent identity/auth\"]}", "in_tok": 530, "out_tok": 203} {"draft_name": "draft-yang-srv6ops-intelligent-routing", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft proposes an intelligent routing method for SR Policy that selects forwarding paths based on network quality metrics in MPLS and IPv6 environments. It extends existing SR Policy candidate path selection with network-aware intelligence for optimized routing decisions.\",\"n\":2,\"nn\":\"Incremental improvement to existing SR Policy mechanisms by adding network quality-based path selection intelligence\",\"m\":3,\"mn\":\"Defines a specific mechanism for intelligent SR Policy routing with clear methodology, though implementation details may need refinement\",\"o\":3,\"on\":\"Shares concepts with existing SR Policy and traffic engineering drafts, builds on established SR framework\",\"mo\":2,\"mon\":\"Single revision draft with unclear working group engagement or broader community interest\",\"r\":3,\"rn\":\"Partially relevant as it involves intelligent/AI-based routing decisions for network optimization, though not core AI agent functionality\",\"c\":[\"ML traffic mgmt\",\"Autonomous netops\"]}", "in_tok": 536, "out_tok": 200} {"draft_name": "draft-yan-anima-brski-cle", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft proposes BRSKI-CLE, a certificateless enrollment framework for constrained IoT devices that cannot handle certificates due to RAM limitations. The approach uses Key Encapsulation Mechanism (KEM) for quantum-safe security and lightweight credentials like CBOR Web Tokens instead of traditional X.509 certificates.\",\"n\":3,\"nn\":\"Useful contribution addressing a real constraint problem in IoT bootstrapping, though certificateless cryptography concepts exist\",\"m\":2,\"mn\":\"Early sketch with framework outline but lacks detailed protocol specification, implementation details, and concrete examples\",\"o\":3,\"on\":\"Shares bootstrapping concepts with existing BRSKI work and builds on established certificateless cryptography approaches\",\"mo\":2,\"mon\":\"Single revision draft with no clear working group adoption or community feedback visible\",\"r\":2,\"rn\":\"Focuses on IoT device bootstrapping and authentication rather than AI agents, though could tangentially apply to constrained AI devices\",\"c\":[\"Agent identity/auth\"]}", "in_tok": 670, "out_tok": 220} {"draft_name": "draft-intellinode-ai-semantic-contract", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"Proposes a semantic-driven contract mechanism where AI applications provide minimum necessary semantics to networks, enabling differentiated forwarding and resource allocation for tensor flows. Aims to improve resource utilization and task completion times in heterogeneous AI computing environments.\",\n \"n\": 4,\n \"nn\": \"Notable approach to expose AI workload semantics to network layer for intelligent traffic shaping, going beyond traditional QoS\",\n \"m\": 2,\n \"mn\": \"Early stage draft at 6 pages, likely contains high-level concepts but lacks detailed protocol specification and implementation details\",\n \"o\": 2,\n \"on\": \"Some conceptual overlap with general network QoS and ML-aware networking, but semantic contract approach appears relatively unique\",\n \"mo\": 2,\n \"mon\": \"Single draft version with recent date suggests early development stage without demonstrated community engagement\",\n \"r\": 4,\n \"rn\": \"Directly addresses AI network optimization challenges, highly relevant to distributed AI training and inference infrastructure\",\n \"c\": [\"ML traffic mgmt\", \"Autonomous netops\"]\n}", "in_tok": 608, "out_tok": 257} {"draft_name": "draft-gaikwad-woa", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"Defines a JSON-based format and HTTP convention for advertising and invoking AI agents uniformly across web hosts. Uses JSON Schema for describing agent inputs/outputs and serves from well-known locations as a primitive for higher-level discovery systems.\",\n \"n\": 2,\n \"nn\": \"Basic HTTP+JSON approach to agent description, incremental application of existing web standards to AI agent context\",\n \"m\": 3,\n \"mn\": \"Defines concrete protocol with JSON Schema usage and HTTP conventions, though lacks comprehensive examples and implementation details\",\n \"o\": 3,\n \"on\": \"Shares conceptual ground with OpenAPI/service description formats and other agent discovery approaches\",\n \"mo\": 2,\n \"mon\": \"Single draft submission with no apparent working group adoption or community discussion yet\",\n \"r\": 4,\n \"rn\": \"Directly addresses core challenge of AI agent discovery and standardized invocation on the web\",\n \"c\": [\"Agent discovery/reg\", \"Data formats/interop\", \"A2A protocols\"]\n}", "in_tok": 560, "out_tok": 247} {"draft_name": "draft-barney-caam", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"CAAM defines a post-discovery authorization layer for AI agents that enforces relationship-based access control and purpose-bound delegation after agent discovery but before execution. It introduces Session Context Objects as JWTs/CWTs with contextual assertion claims to bind delegated authority to specific purposes and trust chains.\",\n \"n\": 4,\n \"nn\": \"Notable originality in combining relationship-based access control with purpose-bound delegation for agent authorization, plus the contextual assertion claim design\",\n \"m\": 4,\n \"mn\": \"Detailed 33-page specification with defined protocol mechanisms, token formats, and integration with multiple existing frameworks\",\n \"o\": 3,\n \"on\": \"Shares authorization concepts with existing identity frameworks but creates novel combinations for agent-specific use cases\",\n \"mo\": 2,\n \"mon\": \"Single revision with complex dependencies on multiple other draft protocols, unclear adoption signals\",\n \"r\": 5,\n \"rn\": \"Core AI agent topic addressing critical authorization and delegation challenges in multi-agent systems\",\n \"c\": [\"Agent identity/auth\", \"A2A protocols\", \"Policy/governance\", \"Human-agent interaction\"]\n}", "in_tok": 842, "out_tok": 273} {"draft_name": "draft-jiang-seat-dynamic-attestation", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"Proposes dynamic attestation mechanisms for AI agent communication over TLS, addressing how to continuously verify agent runtime posture including platform TCB, models, tools and policies during long-lived sessions. Focuses on enabling relying parties to make authorization decisions based on current agent state rather than initial handshake credentials.\",\"n\":4,\"nn\":\"Notable originality in applying continuous attestation to AI agent contexts where runtime state changes frequently\",\"m\":2,\"mn\":\"Early sketch level - describes use case and requirements but lacks detailed protocol mechanisms\",\"o\":2,\"on\":\"Minor similarities to general remote attestation work but AI agent context is relatively unique\",\"mo\":2,\"mon\":\"Single revision with no clear working group adoption yet\",\"r\":5,\"rn\":\"Core AI agent topic addressing fundamental trust and security challenges in agent-to-agent communication\",\"c\":[\"A2A protocols\",\"Agent identity/auth\"]}", "in_tok": 577, "out_tok": 193} {"draft_name": "draft-improving-data-quality-tags", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft proposes using special text tags to improve data quality and context for conversational AI models during training and inference. The approach aims to enhance model understanding of user queries through structured markup.\", \"n\":2,\"nn\":\"Incremental approach to AI training data structuring, building on established markup concepts\",\"m\":2,\"mn\":\"Early stage draft with basic concept outline but lacks detailed protocol specification\",\"o\":3,\"on\":\"Shares general concepts with existing AI training and markup standardization efforts\",\"mo\":2,\"mon\":\"Single revision with no apparent working group adoption or community engagement\",\"r\":4,\"rn\":\"Directly addresses AI model improvement through structured data approaches\",\"c\":[\"Data formats/interop\",\"Model serving/inference\"]}", "in_tok": 505, "out_tok": 158} {"draft_name": "draft-hw-protocol-agent", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft proposes network protocol enhancements to recognize and optimize transmission of multi-modal AI agent data. It addresses the need for networks to understand different quality requirements for various data types (text, audio, video) within AI agent communications.\",\"n\":3,\"nn\":\"Useful contribution addressing a real emerging need, though applying existing QoS concepts to AI agent traffic\",\"m\":2,\"mn\":\"Early sketch phase with problem statement and high-level requirements but lacks detailed protocol specifications\",\"o\":3,\"on\":\"Shares QoS and traffic management concepts with existing network drafts but focuses specifically on AI agent multi-modal requirements\",\"mo\":2,\"mon\":\"Single revision draft from 2026 with no clear working group adoption or community engagement\",\"r\":4,\"rn\":\"Directly relevant to AI agent networking needs, addressing practical infrastructure requirements for multi-modal agent communications\",\"c\":[\"ML traffic mgmt\",\"A2A protocols\"]}", "in_tok": 581, "out_tok": 202} {"draft_name": "draft-zhang-rtgwg-llmmoe-multicast", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft analyzes potential multicast use cases in Large Language Model Mixture of Experts (MoE) architectures, particularly for token dispatching scenarios. The document appears to be a problem statement and use case analysis rather than a detailed protocol specification.\",\"n\":2,\"nn\":\"Incremental work applying existing multicast concepts to MoE token distribution - useful but not particularly original\",\"m\":2,\"mn\":\"Early sketch level - analyzes use cases but lacks detailed protocol mechanisms or specifications\",\"o\":2,\"on\":\"Minor similarities to general ML traffic management approaches but addresses specific MoE multicast scenarios\",\"mo\":2,\"mon\":\"Single revision with future date suggests early stage work without demonstrated community engagement\",\"r\":4,\"rn\":\"Directly relevant to AI model serving and inference optimization, addressing real networking challenges in MoE architectures\",\"c\":[\"Model serving/inference\",\"ML traffic mgmt\"]}", "in_tok": 540, "out_tok": 196} {"draft_name": "draft-yang-spring-sr-policy-intelligent-routing", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft proposes an intelligent routing method for SR Policy that selects paths based on network quality metrics in MPLS and IPv6 environments. It extends existing SR Policy candidate path selection with network-aware decision making capabilities.\",\n \"n\": 2,\n \"nn\": \"Incremental enhancement to existing SR Policy mechanisms by adding network quality-based path selection\",\n \"m\": 3,\n \"mn\": \"Defines a specific protocol mechanism with clear concepts but lacks detailed implementation specifics\",\n \"o\": 3,\n \"on\": \"Shares concepts with existing SR Policy and traffic engineering drafts but adds distinct quality-based routing approach\",\n \"mo\": 2,\n \"mon\": \"Single revision with no clear working group adoption or community feedback visible\",\n \"r\": 3,\n \"rn\": \"Partially relevant as it describes intelligent/adaptive network behavior but focuses on traditional networking rather than AI agents\",\n \"c\": [\"ML traffic mgmt\", \"Autonomous netops\"]\n}", "in_tok": 531, "out_tok": 235} {"draft_name": "draft-ietf-websec-mime-sniff", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft specifies an algorithm for determining effective media types of HTTP responses by examining both Content-Type headers and response content, addressing compatibility issues with web servers that supply incorrect MIME types. The algorithm balances security considerations with the need to handle legacy server behavior that commonly provides mismatched content types.\",\n \"n\": 2,\n \"nn\": \"Incremental standardization of existing browser behavior for MIME type sniffing\",\n \"m\": 4,\n \"mn\": \"Detailed specification with concrete algorithm steps and security considerations\",\n \"o\": 4,\n \"on\": \"Significant overlap with existing web security and HTTP standards\",\n \"mo\": 3,\n \"mon\": \"Active IETF working group development in websec\",\n \"r\": 1,\n \"rn\": \"Pure web security/HTTP protocol specification with no AI or agent relevance\",\n \"c\": [\"Other AI/agent\"]\n}", "in_tok": 543, "out_tok": 221} {"draft_name": "draft-birkholz-verifiable-agent-conversations", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft defines a CDDL-based data format for creating verifiable records of autonomous agent conversations, capturing session metadata, message exchanges, tool invocations, and reasoning traces. It uses COSE signing for integrity and aims to enable accountability and compliance for AI agent interactions through structured, interoperable conversation logs.\",\"n\":3,\"nn\":\"Useful contribution addressing the emerging need for verifiable AI agent interactions, though the core concept of signed conversation logs is relatively straightforward\",\"m\":4,\"mn\":\"Detailed 51-page specification with comprehensive CDDL definitions, JSON/CBOR representations, and clear integration with existing standards like COSE and SCITT\",\"o\":2,\"on\":\"Some conceptual overlap with general verifiable data structures and audit logging, but specific focus on AI agent conversations is relatively unique\",\"mo\":2,\"mon\":\"Single draft submission without clear indication of working group adoption or broader community engagement\",\"r\":5,\"rn\":\"Directly addresses a critical need in AI agent systems for accountability, transparency, and verifiable behavior records as agents perform increasingly consequential tasks\",\"c\":[\"Data formats/interop\",\"AI safety/alignment\",\"A2A protocols\"]}", "in_tok": 672, "out_tok": 256} {"draft_name": "draft-zyyhl-agent-networks-framework", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft defines a framework for AI agent networks using the Agent Network Protocol (ANP) to enable communication between AI agents within trust domains. It appears to establish basic communication functions for agent-to-agent interactions in networked environments.\",\"n\":3,\"nn\":\"Useful contribution defining structured agent communication framework, though agent networking concepts are becoming established\",\"m\":2,\"mn\":\"Early sketch stage with basic framework definition but lacking detailed protocol specifications and implementation details\",\"o\":3,\"on\":\"Shares networking and communication concepts with emerging agent protocol work but focuses specifically on trust domain approach\",\"mo\":2,\"mon\":\"Single revision draft with no clear working group adoption or broader community engagement yet\",\"r\":5,\"rn\":\"Core AI agent topic directly addressing agent-to-agent communication protocols and network architectures\",\"c\":[\"A2A protocols\",\"Agent discovery/reg\",\"Agent identity/auth\"]}", "in_tok": 495, "out_tok": 189} {"draft_name": "draft-nivalto-agentroa-route-authorization", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft specifies AgentROA, a cryptographic policy enforcement framework for AI agents that adapts BGP Route Origin Authorization concepts to govern agent actions through signed policy envelopes, delegation chains, and execution receipts. The framework implements a Border Gateway enforcement model that positions cryptographic validation at capability invocation boundaries, external to the agent's execution context.\",\"n\":4,\"nn\":\"Novel application of RPKI/BGP ROA concepts to AI agent governance with creative Border Gateway enforcement positioning\",\"m\":4,\"mn\":\"Detailed specification with clear protocol objects, schemas, and enforcement semantics across multiple deployment topologies\",\"o\":2,\"on\":\"Unique cryptographic approach to agent governance with minimal overlap to existing AI safety frameworks\",\"mo\":2,\"mon\":\"Single revision draft with no clear working group adoption or community feedback yet\",\"r\":5,\"rn\":\"Directly addresses core AI agent governance and policy enforcement challenges with practical cryptographic mechanisms\",\"c\":[\"Policy/governance\",\"Agent identity/auth\",\"AI safety/alignment\"]}", "in_tok": 879, "out_tok": 220} {"draft_name": "draft-li-spring-rdma-multicast-over-srv6", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft extends SRv6 to enable multicast delivery of RDMA traffic by defining new endpoint behaviors for per-receiver packet modifications and aggregating response messages. It targets high-performance computing and AI training scenarios where efficient one-to-many data distribution is critical.\",\n \"n\": 4,\n \"nn\": \"Notable originality in combining SRv6 segment routing with RDMA multicast - addresses complex challenge of maintaining QP semantics over IP multicast trees\",\n \"m\": 4,\n \"mn\": \"Detailed specification with new endpoint behavior End.MT, BTH modification procedures, and ACK/NACK aggregation mechanisms - appears implementation-ready\",\n \"o\": 2,\n \"on\": \"Minor similarities to general SRv6 and RDMA work but unique in combining these for multicast scenarios\",\n \"mo\": 3,\n \"mon\": \"Active development with recent draft date, targeting emerging HPC/AI infrastructure needs\",\n \"r\": 4,\n \"rn\": \"Directly relevant to AI agent infrastructure - specifically addresses AI training parameter distribution and inference KV cache distribution at scale\",\n \"c\": [\"ML traffic mgmt\", \"Model serving/inference\"]\n}", "in_tok": 675, "out_tok": 286} {"draft_name": "draft-agent-gw", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft proposes an Agent Communication Gateway (Agent-GW) that serves as an infrastructure hub for multi-agent collaboration, providing semantic routing, shared working memory, and protocol adaptation across heterogeneous agent systems. The framework aims to enable intent-based interactions and collaborative workflows between agents across different administrative and protocol boundaries.\",\"n\":4,\"nn\":\"Notable integration of semantic routing with working memory and knowledge delivery networks - combines existing concepts in a genuinely useful way for agent infrastructure\",\"m\":2,\"mn\":\"Early architectural sketch with high-level concepts but lacks detailed protocol specifications, message formats, or concrete implementation guidance\",\"o\":2,\"on\":\"While touches on gateway concepts, the semantic routing and working memory combination for agent infrastructure appears relatively unique\",\"mo\":2,\"mon\":\"Single revision from 2026 with no indicators of working group adoption or broader community engagement\",\"r\":5,\"rn\":\"Directly addresses core challenges in AI agent communication, collaboration, and infrastructure - highly relevant to emerging agent ecosystems\",\"c\":[\"A2A protocols\",\"Agent discovery/reg\",\"Data formats/interop\",\"Model serving/inference\"]}", "in_tok": 622, "out_tok": 238} {"draft_name": "draft-sato-soos-cap", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"CAP defines a three-tier prohibition architecture for agentic AI systems with kernel-enforced constraints, evaluating both AI agent actions and human principal decisions against absolute, jurisdictional, and voluntary ethical prohibitions. The protocol includes jurisdictional conflict resolution mechanisms to handle irreconcilable policy conflicts via human escalation.\",\"n\":4,\"nn\":\"Novel three-tier prohibition model with dual evaluation points and kernel enforcement represents a genuinely original approach to AI governance\",\"m\":4,\"mn\":\"Detailed specification with defined tiers, evaluation mechanisms, and integration with related protocols, approaching implementation readiness\",\"o\":2,\"on\":\"While governance concepts exist, the specific kernel-enforced dual-evaluation architecture has minimal overlap with existing work\",\"mo\":3,\"mon\":\"Active development evidenced by cross-references to related drafts in the same series, suggesting coordinated protocol suite development\",\"r\":5,\"rn\":\"Directly addresses core AI agent safety and governance concerns with technical protocol specification\",\"c\":[\"AI safety/alignment\",\"Policy/governance\",\"Human-agent interaction\"]}", "in_tok": 662, "out_tok": 230} {"draft_name": "draft-dhir-http-agent-profile", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"Proposes HTTP Agent Profile (HAP) to enable cryptographic authentication of autonomous agents, separate human from agent traffic using privacy-preserving tokens, and enable micropayments for agent content access. Addresses economic misalignment between AI systems that consume content at scale without viewing ads and content providers who rely on advertising revenue.\",\n \"n\": 4,\n \"nn\": \"Notable integration of agent authentication, human/agent traffic separation, and micropayment mechanisms into a cohesive HTTP-based protocol addressing a genuine economic problem\",\n \"m\": 3,\n \"mn\": \"Well-defined protocol mechanics with clear use of HTTP Message Signatures and status codes, though lacks detailed implementation examples and test vectors\",\n \"o\": 2,\n \"on\": \"While authentication and payment protocols exist separately, this specific combination for agent traffic management appears to have minimal overlap with existing work\",\n \"mo\": 2,\n \"mon\": \"Recent draft from November 2024 but appears to be initial publication without evidence of working group adoption or broader community engagement\",\n \"r\": 5,\n \"rn\": \"Directly addresses core challenges of AI agent web traffic including authentication, economic sustainability, and human/agent differentiation - highly relevant to AI agent ecosystem\",\n \"c\": [\"Agent identity/auth\", \"ML traffic mgmt\", \"Policy/governance\"]\n}", "in_tok": 686, "out_tok": 309} {"draft_name": "draft-ietf-emu-pqc-eapaka", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft enhances EAP-AKA' Forward Secrecy protocol with Post-Quantum Key Encapsulation Mechanisms to protect against future quantum computer threats. It replaces traditional ECDHE key agreement with quantum-resistant algorithms to maintain perfect forward secrecy in mobile authentication.\",\n \"n\": 2,\n \"nn\": \"Incremental upgrade replacing ECDHE with PQ-KEMs in existing EAP-AKA' FS - straightforward quantum-resistant enhancement\",\n \"m\": 3,\n \"mn\": \"20-page draft with defined protocol mechanisms building on RFC9678, appears to specify concrete PQ-KEM integration\",\n \"o\": 3,\n \"on\": \"Shares post-quantum cryptography concepts with other PQC drafts, builds directly on existing EAP-AKA' FS specification\",\n \"mo\": 4,\n \"mon\": \"IETF EMU working group draft with 2026 date suggests active development and working group adoption\",\n \"r\": 1,\n \"rn\": \"This is about quantum-resistant cryptography for mobile network authentication, not AI or autonomous agents\",\n \"c\": [\"Other AI/agent\"]\n}", "in_tok": 712, "out_tok": 292} {"draft_name": "draft-chen-nmrg-ibn-management", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft proposes an intent-based network management use case with a five-level architecture for achieving fully autonomous network operations without human intervention. It focuses on simplifying network complexity through automated management driven by high-level intent specifications.\",\"n\":2,\"nn\":\"Incremental contribution building on well-established IBN concepts without significant new insights\",\"m\":2,\"mn\":\"Early conceptual sketch with basic architecture outline but lacks detailed protocols or implementation specifics\",\"o\":4,\"on\":\"Significant overlap with existing IBN and autonomous network management literature and standards\",\"mo\":2,\"mon\":\"Single revision with limited evidence of community engagement or working group interest\",\"r\":4,\"rn\":\"Directly relevant to AI-driven network operations and autonomous management systems\",\"c\":[\"Autonomous netops\",\"Policy/governance\"]}", "in_tok": 562, "out_tok": 173} {"draft_name": "draft-steele-agent-considerations", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft proposes adding 'Agent Considerations' sections to IETF specifications to help AI agents consume specs and generate code from them. It defines guidance for authorship, examples annotation, and language-specific media-type production to enable better AI-assisted implementation of IETF protocols.\",\"n\":3,\"nn\":\"Useful contribution addressing the practical need for AI-readable specifications, though the core concept of structured documentation for automated consumption isn't entirely new\",\"m\":2,\"mn\":\"Early sketch level - defines the concept and structure but lacks detailed implementation guidance, specific examples, or concrete mechanisms for how agents would actually consume this information\",\"o\":2,\"on\":\"Minor conceptual similarities to existing documentation standards and API specification formats, but the specific focus on IETF specs for AI agent consumption is relatively unique\",\"mo\":2,\"mon\":\"Single revision with no clear indication of working group interest or broader community engagement in this specific approach\",\"r\":4,\"rn\":\"Directly relevant to AI agents in networking - addresses the practical challenge of enabling AI systems to understand and implement network protocols from IETF specifications\",\"c\":[\"Data formats/interop\",\"Policy/governance\",\"Human-agent interaction\"]}", "in_tok": 604, "out_tok": 257} {"draft_name": "draft-ai-traffic", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft presents a problem statement for handling AI/LLM network traffic across datacenters, focusing on the specific characteristics of training and inference workloads that require federated cooperation between service providers. The document identifies networking challenges arising from large parameter counts and collective communication patterns in distributed AI systems.\",\n \"n\": 2,\n \"nn\": \"Incremental work identifying known networking challenges for distributed AI workloads\",\n \"m\": 1,\n \"mn\": \"Pure problem statement without proposed solutions or mechanisms\",\n \"o\": 3,\n \"on\": \"Shares concepts with existing work on AI traffic optimization and distributed ML networking\",\n \"mo\": 2,\n \"mon\": \"Single revision from October 2024, unclear community engagement\",\n \"r\": 4,\n \"rn\": \"Directly addresses AI infrastructure challenges, though focused on network layer rather than agent protocols\",\n \"c\": [\"ML traffic mgmt\", \"Model serving/inference\"]\n}", "in_tok": 603, "out_tok": 228} {"draft_name": "draft-liu-agent-metadata-sync-protocol", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"Defines AMSP for synchronizing agent metadata between hierarchical gateways in Internet of Agents deployments. Enables global agent discovery and routing across distributed network domains with loop-free propagation.\",\"n\":3,\"nn\":\"Useful contribution applying distributed systems concepts to agent infrastructure, though builds on established gateway synchronization patterns\",\"m\":3,\"mn\":\"Defines core protocol mechanisms and hierarchical architecture but lacks detailed message formats and implementation specifics\",\"o\":2,\"on\":\"Minor similarities to general distributed synchronization protocols but specific application to agent metadata is relatively unique\",\"mo\":2,\"mon\":\"Single revision with future date suggests early stage development, no clear working group adoption\",\"r\":5,\"rn\":\"Core AI agent infrastructure topic addressing fundamental challenges of large-scale agent deployments and inter-gateway coordination\",\"c\":[\"A2A protocols\",\"Agent discovery/reg\",\"Autonomous netops\"]}", "in_tok": 595, "out_tok": 193} {"draft_name": "draft-shang-campus-agent-scope-down", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft identifies the authorization challenges when AI agents operate autonomously in enterprise campus networks, proposing task-bound privilege reduction enforced by network devices. It argues for in-path enforcement to maintain compatibility with legacy systems while controlling agent access across heterogeneous protocols.\",\n \"n\": 4,\n \"nn\": \"Novel application of network-based enforcement for AI agent authorization in campus environments, addressing the mismatch between traditional human-centric auth models and autonomous agent behavior\",\n \"m\": 2,\n \"mn\": \"Early problem statement and high-level solution sketch, lacks concrete protocol definitions or implementation details\",\n \"o\": 2,\n \"on\": \"Some conceptual overlap with zero-trust architectures but unique focus on AI agent scope-down in campus networks\",\n \"mo\": 2,\n \"mon\": \"Single draft submission with no indication of working group adoption or broader community engagement\",\n \"r\": 5,\n \"rn\": \"Directly addresses core AI agent authentication and authorization challenges in network environments\",\n \"c\": [\"Agent identity/auth\", \"Policy/governance\", \"Autonomous netops\"]\n}", "in_tok": 608, "out_tok": 261} {"draft_name": "draft-pbs-sidrops-roaanycast", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft provides best practices for managing Route Origin Authorizations (ROAs) in RPKI when using anycast services where each anycast node has a unique origin ASN. It focuses on mitigating technical risks in RPKI operations for such deployments.\",\"n\":2,\"nn\":\"incremental contribution addressing specific operational scenarios in existing RPKI framework\",\"m\":2,\"mn\":\"early sketch of best practices without detailed implementation specifics\",\"o\":3,\"on\":\"builds on existing RPKI and anycast concepts but addresses specific combination\",\"mo\":2,\"mon\":\"single revision with no clear working group adoption signals\",\"r\":1,\"rn\":\"purely network infrastructure focused, no AI or agent relevance\",\"c\":[\"Other AI/agent\"]}", "in_tok": 542, "out_tok": 164} {"draft_name": "draft-sharif-aeba", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"AEBA defines a framework for monitoring and analyzing behavioral events from autonomous AI agents, providing structured event schemas, signature binding, and integration with existing SIEM systems. It aims to enable security monitoring of AI agents similar to how UEBA systems monitor user behavior in enterprise environments.\",\"n\":4,\"nn\":\"Notable originality in adapting behavioral analytics concepts specifically for autonomous AI agents with cryptographic bindings\",\"m\":4,\"mn\":\"Detailed specification with event schemas, protocols, signature mechanisms, and SIEM integration examples\",\"o\":2,\"on\":\"Minor similarities to general behavioral monitoring approaches but unique focus on autonomous AI agents\",\"mo\":2,\"mn\":\"Single revision with no clear working group adoption or community engagement yet\",\"r\":5,\"rn\":\"Core AI agent topic directly addressing critical security monitoring needs for autonomous systems\",\"c\":[\"AI safety/alignment\",\"Agent identity/auth\",\"Data formats/interop\",\"A2A protocols\"]}", "in_tok": 671, "out_tok": 201} {"draft_name": "draft-nandakumar-a2a-moqt-transport", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft specifies how the Agent-to-Agent (A2A) protocol can be transported over Media over QUIC Transport (MOQT), enabling real-time communication between AI agents. It leverages MOQT's publish/subscribe model and streaming capabilities to support A2A requirements for agent discovery, negotiation, and collaboration.\",\"n\":2,\"nn\":\"Incremental application of existing MOQT transport to A2A protocol - straightforward protocol mapping\",\"m\":3,\"mn\":\"Defines specific protocol mechanisms for A2A over MOQT with reasonable detail at 12 pages\",\"o\":3,\"on\":\"Shares transport concepts with other QUIC-based approaches and overlaps with general A2A protocol work\",\"mo\":2,\"mon\":\"Single revision with 2026 date suggests early stage development without established momentum\",\"r\":4,\"rn\":\"Directly addresses AI agent communication transport, core to agent-to-agent interaction infrastructure\",\"c\":[\"A2A protocols\",\"Data formats/interop\"]}", "in_tok": 567, "out_tok": 228} {"draft_name": "draft-barnes-hpke-hpke", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This IETF draft defines HPKE (Hybrid Public Key Encryption), a cryptographic scheme that combines asymmetric and symmetric encryption for efficient public key encryption of arbitrary-sized plaintexts. The scheme includes authenticated variants and works with various combinations of KEMs, KDFs, and AEAD functions, providing instantiations using widely-used primitives like ECDH and HKDF.\",\"n\":3,\"nn\":\"Useful contribution that systematically combines existing cryptographic primitives into a cohesive hybrid encryption scheme with multiple authentication modes\",\"m\":5,\"mn\":\"Implementation-ready specification at 125 pages with detailed protocol definitions, likely includes comprehensive examples and test vectors\",\"o\":2,\"on\":\"While building on established cryptographic primitives, the specific hybrid approach and authentication variants have minor similarities to existing schemes\",\"mo\":4,\"mon\":\"CFRG product suggests working group adoption and active standardization process within IRTF crypto community\",\"r\":1,\"rn\":\"This is a pure cryptographic protocol specification with no AI or agent-related content - appears to be a false positive in AI/agent classification\",\"c\":[\"Other AI/agent\"]}", "in_tok": 688, "out_tok": 257} {"draft_name": "draft-vandemeent-jis-identity", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"JIS defines a comprehensive identity and trust protocol for autonomous agents using dual keypairs (human-device and device authentication), trust negotiation handshakes, and human-readable context layers. The protocol operates transport-agnostically as a semantic layer and integrates with multiple referenced systems for provenance, verification, and agent discovery.\",\n \"n\": 4,\n \"nn\": \"Notable originality in combining dual-keypair identity model with intent binding and human-readable context layers - goes beyond traditional PKI approaches\",\n \"m\": 3,\n \"mn\": \"Defines core protocol mechanisms and handshakes but lacks detailed implementation specifics and test vectors for full deployment readiness\",\n \"o\": 2,\n \"on\": \"While identity and trust are common themes, the specific dual-keypair approach with intent binding and Humotica context layer appears relatively unique\",\n \"mo\": 1,\n \"mn\": \"Future-dated draft (2026) with no indication of working group adoption or community development activity\",\n \"r\": 5,\n \"rn\": \"Directly addresses core autonomous agent challenges of identity establishment, trust negotiation, and intent binding - fundamental requirements for agent systems\",\n \"c\": [\"Agent identity/auth\", \"A2A protocols\", \"Human-agent interaction\"]\n}", "in_tok": 665, "out_tok": 299} {"draft_name": "draft-johani-dnsop-transport-signaling", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft proposes a mechanism for authoritative DNS servers to signal support for encrypted transport protocols (DoT, DoH, DoQ) within DNS responses or through direct queries. It enables resolvers to discover and upgrade to secure transports opportunistically or with DNSSEC validation, improving upon RFC9539's blind transport testing approach.\",\n \"n\": 2,\n \"nn\": \"Incremental improvement to existing DNS transport discovery mechanisms, building on RFC9539 with signaling rather than blind probing\",\n \"m\": 4,\n \"mn\": \"Detailed specification with clear protocol mechanisms, examples of opportunistic and validated modes, and implementation guidance\",\n \"o\": 3,\n \"on\": \"Shares concepts with existing DNS transport upgrade mechanisms and RFC9539, but provides a distinct signaling-based approach\",\n \"mo\": 3,\n \"mon\": \"Active individual draft with GitHub collaboration, but no clear working group adoption or broader community momentum yet\",\n \"r\": 1,\n \"rn\": \"Pure DNS protocol enhancement for transport discovery - not related to AI agents or autonomous systems\",\n \"c\": [\"Other AI/agent\"]\n}", "in_tok": 899, "out_tok": 269} {"draft_name": "draft-cui-ai-agent-discovery-invocation", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft proposes a standardized protocol for discovering and invoking AI agents across platforms, defining metadata formats, capability-based discovery, and RESTful invocation interfaces. It includes an optional intent-based agent selection extension and addresses security considerations for multi-agent system interoperability.\",\n \"n\": 3,\n \"nn\": \"Useful contribution addressing real interoperability needs, though the core concepts of service discovery and RESTful invocation are well-established patterns being applied to AI agents\",\n \"m\": 3,\n \"mn\": \"Defines concrete protocol mechanisms with metadata schemas and RESTful interfaces, though lacks detailed implementation examples or comprehensive security specifications\",\n \"o\": 3,\n \"on\": \"Shares service discovery concepts with existing web service standards and overlaps with general agent communication protocols, but focuses specifically on AI agent interoperability\",\n \"mo\": 2,\n \"mn\": \"Single revision from future date (2026-02-12) suggests early stage work without demonstrated community engagement or working group adoption\",\n \"r\": 5,\n \"rn\": \"Directly addresses core AI agent infrastructure needs for discovery, metadata description, and cross-platform invocation - fundamental to multi-agent system development\",\n \"c\": [\"Agent discovery/reg\", \"A2A protocols\", \"Agent identity/auth\", \"Data formats/interop\"]\n}", "in_tok": 657, "out_tok": 313} {"draft_name": "draft-yusef-lamps-rfc7030-renewal-recommendation", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft extends RFC 7030 (Enrollment over Secure Transport) to allow certificate authorities to provide specific renewal timing recommendations to end-entity devices, replacing the typical 50% lifetime rule. The motivation is that as certificate lifetimes shorten from years to months, the traditional 50% renewal point may be unnecessarily early and wasteful.\",\"n\":2,\"nn\":\"Incremental extension addressing a practical timing issue as certificate lifetimes decrease\",\"m\":2,\"mn\":\"Early sketch level - only provides problem statement and basic concept without detailed protocol mechanisms\",\"o\":3,\"on\":\"Shares certificate management concepts with existing PKI and EST work but addresses specific renewal timing gap\",\"mo\":2,\"mon\":\"Single revision with recent date suggests limited development activity so far\",\"r\":2,\"rn\":\"Tangentially related to AI/agents only in that automated systems need certificate management, but primarily a general PKI improvement\",\"c\":[\"Agent identity/auth\"]}", "in_tok": 575, "out_tok": 208} {"draft_name": "draft-ar-emu-hybrid-pqc-eapaka", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft proposes enhancing EAP-AKA' Forward Secrecy with hybrid post-quantum cryptography to protect against future quantum computer threats. It aims to make the authentication protocol quantum-safe by replacing traditional ECDHE key exchange with PQ/T hybrid algorithms.\",\n \"n\": 2,\n \"nn\": \"Incremental application of established hybrid PQC concepts to existing EAP-AKA' protocol\",\n \"m\": 2,\n \"mn\": \"Early stage draft with basic problem statement and approach outlined but lacking detailed protocol specification\",\n \"o\": 4,\n \"on\": \"Significant overlap with general hybrid PQC standardization efforts and other quantum-safe authentication proposals\",\n \"mo\": 3,\n \"mon\": \"Active development addressing timely quantum threat concerns in authentication protocols\",\n \"r\": 1,\n \"rn\": \"This is a network security/cryptography draft about authentication protocols, not related to AI or agents\",\n \"c\": []\n}", "in_tok": 722, "out_tok": 241} {"draft_name": "draft-melnikov-sasl2", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft defines SASL2, an extensible authentication and security layer framework that allows protocols to use pluggable authentication mechanisms. It extends the original SASL to support additional authentication tasks like two-factor authentication and password changes.\",\"n\":2,\"nn\":\"Incremental improvement to existing SASL framework with some useful extensions\",\"m\":4,\"mn\":\"Well-defined protocol specification with clear structure and detailed mechanisms\",\"o\":4,\"on\":\"Significant overlap with existing SASL RFCs, though adds new capabilities\",\"mo\":3,\"mon\":\"Active development as indicated by 2025 date, but momentum unclear without WG context\",\"r\":2,\"rn\":\"Authentication framework is tangentially related to AI agents but not AI-specific\",\"c\":[\"Agent identity/auth\"]}", "in_tok": 620, "out_tok": 173} {"draft_name": "draft-ietf-savnet-inter-domain-architecture", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"Proposes an inter-domain SAVNET architecture for AS-level Source Address Validation using both SAV-specific and general information sharing between autonomous systems. Defines architectural components and framework for generating more accurate SAV rules compared to existing approaches.\",\"n\":3,\"nn\":\"Useful contribution extending SAV concepts to inter-domain scenarios with novel information sharing approach\",\"m\":2,\"mn\":\"Early architectural sketch with components defined but lacks detailed protocol specifications\",\"o\":3,\"on\":\"Shares concepts with existing SAV and BGP security work but focuses on inter-domain coordination\",\"mo\":4,\"mon\":\"Active IETF working group draft with clear development trajectory\",\"r\":1,\"rn\":\"Network security architecture focused on source address validation, not related to AI or agents\",\"c\":[\"Other AI/agent\"]}", "in_tok": 657, "out_tok": 177} {"draft_name": "draft-stephan-ai-agent-6g", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft explores AI agent communication protocols for 6G networks, building on 3GPP TR 22.870 use cases and service requirements. It extrapolates potential protocol requirements for agent-to-agent and agent-to-tool communication in future 6G systems.\",\n \"n\": 2,\n \"nn\": \"Incremental work that applies existing AI agent communication concepts to the 6G context, following established 3GPP requirements analysis patterns\",\n \"m\": 2,\n \"mn\": \"Early sketch that primarily references 3GPP use cases and extrapolates high-level protocol requirements without detailed technical specifications\",\n \"o\": 3,\n \"on\": \"Shares conceptual overlap with general AI agent communication work but focuses specifically on 6G network integration\",\n \"mo\": 2,\n \"mon\": \"Single revision draft with no clear indication of working group adoption or broader community engagement in IETF context\",\n \"r\": 4,\n \"rn\": \"Directly relevant to AI agent protocols, though primarily focused on telecommunications network integration rather than general internet protocols\",\n \"c\": [\"A2A protocols\", \"Data formats/interop\", \"ML traffic mgmt\"]\n}", "in_tok": 544, "out_tok": 282} {"draft_name": "draft-singla-agent-identity-protocol", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft proposes a decentralized identity protocol for AI agents using W3C DIDs and capability-based authorization to enable secure multi-agent workflows. It combines cryptographic delegation chains with deterministic validation to eliminate dependence on centralized identity providers.\",\n \"n\": 4,\n \"nn\": \"Notable integration of DIDs with AI agent delegation - first comprehensive framework addressing agent identity chains and capability inheritance in decentralized systems\",\n \"m\": 4,\n \"mn\": \"Detailed 225-page specification suggests comprehensive protocol definition with examples, though implementation readiness depends on test vector completeness\",\n \"o\": 2,\n \"on\": \"Leverages existing W3C DID standards but applies them uniquely to AI agent contexts with novel delegation mechanisms\",\n \"mo\": 2,\n \"mon\": \"Single draft from 2026 with no visible working group adoption yet, though the comprehensive length suggests serious development effort\",\n \"r\": 5,\n \"rn\": \"Directly addresses core AI agent infrastructure need for decentralized identity and authorization - fundamental building block for multi-agent systems\",\n \"c\": [\"Agent identity/auth\", \"A2A protocols\", \"AI safety/alignment\"]\n}", "in_tok": 540, "out_tok": 282} {"draft_name": "draft-li-trustworthy-routing-discovery", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"Proposes a network architecture for trustworthy routing where data is transmitted only through trusted devices, extending BGP and BGPsec protocols to carry trustworthy routing information (TRI). Aims to address security and privacy requirements for high-security applications by enabling route selection based on device trustworthiness.\",\"n\":2,\"nn\":\"Incremental extension of existing BGP/BGPsec protocols to include trust attestation - builds on established concepts but applies them to routing decisions\",\"m\":2,\"mn\":\"Early sketch level - describes architecture and TRI concept but lacks detailed protocol mechanics, message formats, or trust evaluation algorithms\",\"o\":4,\"on\":\"Significant overlap with existing trusted computing, BGPsec extensions, and secure routing literature - trust attestation in networking is well-explored\",\"mo\":2,\"mon\":\"Single revision draft with no clear working group adoption or broader community engagement visible\",\"r\":2,\"rn\":\"Tangentially related to AI/agents - while trustworthy routing could support AI agent communications, this is primarily a general networking security mechanism\",\"c\":[\"Other AI/agent\"]}", "in_tok": 552, "out_tok": 240} {"draft_name": "draft-zeng-opsawg-llm-netconf-gap", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft analyzes gaps in existing network configuration protocols (NETCONF, RESTCONF, gNMI, etc.) for supporting LLM-driven intent-based networking with natural language interfaces. It systematically identifies where current south-bound protocols fall short for conversational, multi-agent orchestration scenarios.\",\n \"n\": 3,\n \"nn\": \"Useful systematic analysis of protocol gaps for LLM integration, though the concept of intent-based networking gaps is somewhat predictable\",\n \"m\": 2,\n \"mn\": \"Early analysis stage - identifies problems and extension points but lacks concrete protocol modifications or detailed solutions\",\n \"o\": 2,\n \"on\": \"Some conceptual overlap with intent-based networking work, but specific focus on LLM protocol gaps is relatively unique\",\n \"mo\": 2,\n \"mon\": \"Single revision with future date suggests early stage work, unclear community engagement or WG interest\",\n \"r\": 4,\n \"rn\": \"Directly addresses AI agent networking with focus on protocol adaptations for LLM-driven operations\",\n \"c\": [\"Autonomous netops\", \"A2A protocols\", \"Data formats/interop\", \"Human-agent interaction\"]\n}", "in_tok": 562, "out_tok": 285} {"draft_name": "draft-chang-agent-token-efficient", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft proposes ADOL (Agentic Data Optimization Layer), a data layer designed to reduce token consumption in agent-to-agent communication by eliminating redundant schema definitions, verbose structures, and inefficient capability exchanges. The approach uses techniques like JSON references, adaptive field inclusion, and controllable response verbosity to optimize context window usage for LLM-based agents.\",\n \"n\": 3,\n \"nn\": \"Addresses a real optimization problem in agent communication with practical techniques, though the individual methods (JSON refs, adaptive fields) are incremental improvements rather than fundamental innovations\",\n \"m\": 2,\n \"mn\": \"At 12 pages, likely contains problem analysis and basic mechanisms but probably lacks the detailed protocol specifications, message formats, and implementation guidance needed for a mature standard\",\n \"o\": 3,\n \"on\": \"Builds upon existing agent communication protocols like MCP and A2A, sharing optimization concepts with other efficiency-focused drafts but targeting the specific domain of token efficiency\",\n \"mo\": 1,\n \"mn\": \"As a single draft from 2026-01-12 with no indication of working group adoption or community engagement, shows minimal momentum in the standardization process\",\n \"r\": 5,\n \"rn\": \"Directly addresses a core challenge in AI agent systems - the efficient use of limited context windows in LLM-based agent communication, which is fundamental to practical agent deployment\",\n \"c\": [\"A2A protocols\", \"Data formats/interop\"]\n}", "in_tok": 710, "out_tok": 349} {"draft_name": "draft-tsyrulnikov-rats-attested-inference-receipt", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft defines Attested Inference Receipt (AIR), a COSE/CWT-based format for creating cryptographically signed receipts that prove confidential AI inference events occurred within hardware-isolated Trusted Execution Environments. The receipts bind model identity, input/output hashes, and attestation metadata into verifiable artifacts for third-party validation.\",\n \"n\": 4,\n \"nn\": \"Notable originality in creating a standardized receipt format specifically for confidential AI inference with TEE integration - addresses the emerging need for verifiable AI computation proofs\",\n \"m\": 4,\n \"mn\": \"Detailed specification with concrete COSE/CWT profiling, specific platform support (AWS Nitro, Intel TDX), and clear technical implementation guidance at 28 pages\",\n \"o\": 2,\n \"on\": \"Minor similarities to general attestation frameworks but uniquely focused on AI inference receipts with specific cryptographic binding of model and computation metadata\",\n \"mo\": 2,\n \"mn\": \"Single revision draft with no clear indication of working group adoption or broader community engagement yet\",\n \"r\": 5,\n \"rn\": \"Core AI agent topic addressing critical needs for verifiable, confidential AI inference - directly relevant to trustworthy AI deployment and verification\",\n \"c\": [\"Agent identity/auth\", \"Model serving/inference\", \"Data formats/interop\"]\n}", "in_tok": 706, "out_tok": 327} {"draft_name": "draft-wiethuechter-drip-det-tada", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft standardizes DRIP Entity Tags (DETs) as identifiers for drone remote identification, enabling trustworthy air domain awareness for unmanned aircraft systems. DETs serve multiple roles including session IDs for privacy, authentication key IDs for accountability, and encryption key IDs for confidentiality in UAS traffic management.\",\n \"n\": 2,\n \"nn\": \"Incremental application of existing DRIP concepts to air domain awareness - extends known identification protocols to UAS context\",\n \"m\": 4,\n \"mn\": \"Detailed 51-page specification with comprehensive registration and usage procedures for DET implementation in UAS systems\",\n \"o\": 3,\n \"on\": \"Shares identification and authentication concepts with other DRIP drafts and general identity management protocols\",\n \"mo\": 3,\n \"mon\": \"Active development with 2026 date indicating ongoing work, though momentum unclear from single draft snapshot\",\n \"r\": 2,\n \"rn\": \"Primarily about drone identification systems rather than AI agents, though UAS may incorporate autonomous capabilities\",\n \"c\": [\"Agent identity/auth\", \"Policy/governance\"]\n}", "in_tok": 573, "out_tok": 271} {"draft_name": "draft-vandemeent-iddrop", "model": "claude-sonnet-4-20250514", "raw": "```json\n{\n \"s\": \"IDDrop defines a protocol for transferring identity claims between human and machine environments using offer-first and request-first modes. It emphasizes sealed carrier truth and semantic classification rather than filename-based identity verification.\",\n \"n\": 4,\n \"nn\": \"Notable approach to identity transfer that moves beyond traditional filename-based verification to sealed carrier truth and causal validation\",\n \"m\": 3,\n \"mn\": \"Appears to define core protocol mechanisms and operating modes but lacks implementation details and test vectors\",\n \"o\": 2,\n \"on\": \"While identity protocols exist, the specific focus on human-machine identity transfer with sealed carriers is relatively unique\",\n \"mo\": 1,\n \"mon\": \"Single draft from future date (2026-05-18) suggests this is either hypothetical or very early stage work\",\n \"r\": 4,\n \"rn\": \"Directly addresses agent identity and authentication challenges, particularly relevant for human-agent interaction scenarios\",\n \"c\": [\"Agent identity/auth\", \"Human-agent interaction\", \"A2A protocols\"]\n}\n```", "in_tok": 622, "out_tok": 258} {"draft_name": "draft-sandowicz-httpbis-httpa2", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"HTTPA/2 extends HTTP to enable trusted communication with Trusted Execution Environments (TEEs) through remote attestation, eliminating dependence on TLS. The protocol treats TEEs as a new type of internet resource that can be attested and communicated with securely.\",\n \"n\": 4,\n \"nn\": \"Notable approach to integrating TEE attestation directly into HTTP protocol layer, moving beyond traditional TLS-based security models\",\n \"m\": 3,\n \"mn\": \"Defines protocol mechanisms and transaction flows but lacks detailed implementation examples and test vectors\",\n \"o\": 2,\n \"on\": \"Limited overlap with existing work - builds on original HTTPA/1 but represents distinct architectural approach\",\n \"mo\": 2,\n \"mon\": \"Single author draft with no apparent working group adoption or community discussion\",\n \"r\": 3,\n \"rn\": \"Partially relevant - TEE attestation mechanisms could support secure AI agent execution and verification\",\n \"c\": [\"Agent identity/auth\", \"AI safety/alignment\"]\n}", "in_tok": 715, "out_tok": 253} {"draft_name": "draft-meta-layer-overview", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"Proposes a Meta-layer as a coordination substrate enabling communities and agents to share presence, annotations, and governance across web platforms. The draft outlines high-level concepts and use cases but lacks concrete protocol specifications or implementation details.\",\n \"n\": 3,\n \"nn\": \"Useful contribution combining existing concepts (overlay networks, annotation systems, governance) into a coherent coordination framework for multi-agent environments\",\n \"m\": 2,\n \"mn\": \"Early conceptual sketch with terminology and use cases but missing concrete protocols, message formats, or implementation specifications\",\n \"o\": 3,\n \"on\": \"Shares concepts with existing annotation systems, overlay networks, and distributed governance proposals but applies them to a specific coordination use case\",\n \"mo\": 2,\n \"mon\": \"Single revision with future date (2025-10-06), no evidence of working group adoption or community development activity\",\n \"r\": 4,\n \"rn\": \"Directly relevant to AI agents with explicit focus on agent coordination, multi-agent governance, and shared digital spaces for automated systems\",\n \"c\": [\"A2A protocols\", \"Policy/governance\", \"Agent discovery/reg\"]\n}", "in_tok": 586, "out_tok": 275} {"draft_name": "draft-akhavain-moussa-ai-network", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"Proposes DA-ITN, a unified multi-plane network framework to support AI training, inference, and agentic interactions across distributed resources. The architecture includes control, data, and operations planes connecting AI clients, data providers, model providers, and computational resources.\",\n \"n\": 3,\n \"nn\": \"Useful contribution that combines existing networking concepts with AI-specific requirements, though the multi-plane architecture approach is incremental rather than groundbreaking\",\n \"m\": 2,\n \"mn\": \"Early sketch stage with high-level architecture description but lacks detailed protocol specifications, concrete mechanisms, or implementation guidance\",\n \"o\": 3,\n \"on\": \"Shares multi-plane networking concepts with existing IETF work and overlaps with other AI infrastructure proposals, but applies them specifically to AI workflows\",\n \"mo\": 1,\n \"mon\": \"Future-dated draft (2026) suggests this is inactive or hypothetical work with no apparent community engagement or development history\",\n \"r\": 4,\n \"rn\": \"Directly addresses core AI infrastructure challenges including training, inference, and agent interactions with network-level considerations highly relevant to IETF\",\n \"c\": [\"ML traffic mgmt\", \"Model serving/inference\", \"Agent discovery/reg\", \"Data formats/interop\", \"Policy/governance\"]\n}", "in_tok": 763, "out_tok": 309} {"draft_name": "draft-chuyi-nmrg-ai-agent-network", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft explores how large language models and AI agents can enhance network operations and maintenance by reducing manual tasks and improving automation. It identifies key scenarios where AI can address operational challenges and examines the impacts on network architectures, data handling, and standardization needs.\",\n \"n\": 2,\n \"nn\": \"Incremental application of well-known LLM capabilities to network operations domain\",\n \"m\": 2,\n \"mn\": \"Early conceptual sketch identifying scenarios and challenges without concrete protocols or mechanisms\",\n \"o\": 3,\n \"on\": \"Shares general concepts with other network automation and AI operations drafts but focuses specifically on LLM agents\",\n \"mo\": 2,\n \"mon\": \"Single draft with future date suggests early stage work without demonstrated community engagement\",\n \"r\": 4,\n \"rn\": \"Directly addresses AI agents in networking context with specific focus on operational applications\",\n \"c\": [\"Autonomous netops\", \"Policy/governance\", \"Human-agent interaction\"]\n}", "in_tok": 618, "out_tok": 239} {"draft_name": "draft-romanchuk-normative-admissibility", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"Defines a framework for evaluating whether speech acts from autonomous agents are admissible based on modality and grounding state, without requiring semantic truth verification. Enables deterministic auditing of agent communications through rule-based admissibility criteria.\",\n \"n\": 4,\n \"nn\": \"Novel approach to agent communication evaluation focusing on structural admissibility rather than semantic truth - addresses a gap in formal frameworks for agent speech act validation\",\n \"m\": 3,\n \"mn\": \"Defines clear framework with modality categories and grounding states, but lacks detailed implementation specifics and concrete protocol mechanisms\",\n \"o\": 2,\n \"on\": \"Some conceptual overlap with general agent communication work, but the specific focus on normative admissibility evaluation is relatively unique\",\n \"mo\": 2,\n \"mn\": \"Single revision with no clear working group adoption or community engagement signals\",\n \"r\": 4,\n \"rn\": \"Directly addresses autonomous agent communication validation, a core concern for AI agent systems deployment and auditing\",\n \"c\": [\"A2A protocols\", \"AI safety/alignment\", \"Policy/governance\"]\n}", "in_tok": 558, "out_tok": 266} {"draft_name": "draft-somoza-atn-agent-trust-negotiation", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"Defines the Agent Trust Negotiation (ATN) protocol that enables AI agents to establish verifiable working agreements by binding capability statements, delegation proofs, provenance records, and reputation metrics to DNS-discovered agent identities. ATN addresses the critical gap between agent discovery and establishing trusted operational relationships.\",\n \"n\": 4,\n \"nn\": \"Novel approach to binding multiple trust artifacts (capability, delegation, provenance, reputation) into a cohesive negotiation protocol for AI agents\",\n \"m\": 4,\n \"mn\": \"48-page detailed specification with defined protocol mechanisms and comprehensive framework for agent trust establishment\",\n \"o\": 2,\n \"on\": \"While it touches on identity and discovery, the specific trust negotiation binding approach appears largely unique\",\n \"mo\": 1,\n \"mon\": \"Single draft from 2026 with no indication of working group adoption or community development\",\n \"r\": 5,\n \"rn\": \"Directly addresses core AI agent challenges of trust, authorization, and secure inter-agent collaboration\",\n \"c\": [\"A2A protocols\", \"Agent identity/auth\", \"Policy/governance\"]\n}", "in_tok": 549, "out_tok": 269} {"draft_name": "draft-vesely-dmarc-mlm-transform", "model": "claude-sonnet-4-20250514", "raw": "```json\n{\n \"s\": \"This draft proposes a method for mailing list managers to revert From: header transformations that were implemented as DMARC workarounds. It focuses on an author-centric strategy where the signing agent enables undoing of From: rewriting to restore original headers after message reception.\",\n \"n\": 2,\n \"nn\": \"Incremental solution addressing known DMARC/MLM interoperability issues with straightforward header restoration approach\",\n \"m\": 3,\n \"mn\": \"Defines a specific protocol mechanism for undoing MLM transformations with clear author-centric strategy\",\n \"o\": 4,\n \"on\": \"Significant overlap with existing DMARC mitigation techniques and MLM interoperability solutions in email standards\",\n \"mo\": 2,\n \"mon\": \"Single revision draft with no clear indication of working group adoption or broader community engagement\",\n \"r\": 1,\n \"rn\": \"Pure email protocol engineering focused on DMARC/MLM interoperability - not related to AI or autonomous agents\",\n \"c\": [\"Other AI/agent\"]\n}\n```", "in_tok": 591, "out_tok": 268} {"draft_name": "draft-nyantakyi-vaip-agent-identity", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"VAIP defines a cryptographic identity protocol specifically for autonomous AI agents, addressing gaps in existing human-centric identity systems. The protocol provides verifiable identity, hierarchical permissions, trust scoring, and audit trails using Ed25519 signatures and SHA-256 hashing.\",\n \"n\": 4,\n \"nn\": \"Notable originality in designing identity primitives specifically for AI agents rather than adapting human identity protocols\",\n \"m\": 3,\n \"mn\": \"Defines core protocol mechanisms and permission structures but lacks implementation details and test vectors\",\n \"o\": 2,\n \"on\": \"Minor conceptual overlap with general identity protocols but unique focus on AI agent-specific requirements\",\n \"mo\": 2,\n \"mon\": \"Single draft with no indication of working group adoption or community feedback\",\n \"r\": 5,\n \"rn\": \"Directly addresses core AI agent infrastructure needs for identity and authorization in autonomous systems\",\n \"c\": [\"Agent identity/auth\", \"Policy/governance\", \"A2A protocols\"]\n}", "in_tok": 615, "out_tok": 245} {"draft_name": "draft-wu-sidr-aspa-validation-signaling", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft defines a BGP extended community to signal AS_PATH validation state based on ASPA within an autonomous system. It allows IBGP speakers to receive validation state information and use it in local policy decisions for routing.\",\"n\":2,\"nn\":\"Incremental extension adding signaling for existing ASPA validation - useful but straightforward application of known patterns\",\"m\":3,\"mn\":\"Defines the extended community format and basic usage but lacks detailed implementation examples and comprehensive error handling\",\"o\":3,\"on\":\"Shares concepts with other BGP signaling extensions and ASPA-related drafts, building on established BGP extended community mechanisms\",\"mo\":2,\"mon\":\"Recent draft but appears to be early single revision without clear working group adoption signals\",\"r\":1,\"rn\":\"This is purely about BGP routing protocol extensions for AS path validation - not related to AI or autonomous agents despite 'autonomous system' networking terminology\",\"c\":[\"Other AI/agent\"]}", "in_tok": 523, "out_tok": 210} {"draft_name": "draft-liu-dmsc-gw-requirements", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft defines requirements for gateways to support Dynamic Multi-agents Secured Collaboration (DMSC), focusing on scalability, communication efficiency, and security. It identifies gaps in current gateway solutions and proposes the need for specialized DMSC gateways.\",\"n\":2,\"nn\":\"Incremental work applying existing gateway concepts to multi-agent systems without novel architectural insights\",\"m\":2,\"mn\":\"Early requirements document without detailed technical specifications or concrete protocols\",\"o\":3,\"on\":\"Shares gateway and multi-agent communication concepts with existing networking and distributed systems work\",\"mo\":2,\"mon\":\"Single revision draft with no clear working group activity or community engagement\",\"r\":4,\"rn\":\"Directly addresses AI agent collaboration infrastructure, a core concern for multi-agent systems\",\"c\":[\"A2A protocols\",\"Agent identity/auth\",\"Data formats/interop\"]}", "in_tok": 533, "out_tok": 189} {"draft_name": "draft-nemethi-aid-agent-identity-discovery", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"AID defines a DNS-based discovery protocol for locating agent service endpoints using TXT records at _agent. subdomains. The protocol provides a minimal mechanism to discover agent URIs, protocols, and authentication hints before delegating to specialized protocols like MCP or A2A.\",\n \"n\": 2,\n \"nn\": \"Incremental application of well-established DNS service discovery patterns to agent endpoints, similar to existing SRV/TXT record approaches\",\n \"m\": 4,\n \"mn\": \"Detailed specification with clear record format, discovery algorithm, security requirements, and IANA registrations ready for implementation\",\n \"o\": 3,\n \"on\": \"Shares DNS-based discovery concepts with general service discovery mechanisms, but agent-specific application has limited direct overlap\",\n \"mo\": 2,\n \"mon\": \"Single revision with future date suggests early stage, no evidence of working group adoption or broader community engagement yet\",\n \"r\": 4,\n \"rn\": \"Directly addresses agent infrastructure needs for service discovery, essential for agent ecosystem interoperability and deployment\",\n \"c\": [\"Agent discovery/reg\", \"Agent identity/auth\", \"Data formats/interop\"]\n}", "in_tok": 620, "out_tok": 283} {"draft_name": "draft-stone-adrp", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"ADRP defines a dispute resolution protocol for agent-to-agent transactions, separating cryptographic proof validation from contractual satisfaction through arbitration mandates. It introduces counter-attestation overrides and legally-grounded arbitration mechanisms for contested AI agent transactions.\",\n \"n\": 4,\n \"nn\": \"Novel separation of cryptographic proof from contractual satisfaction in agent commerce, introducing arbitration mandates as fourth element of mandate chain\",\n \"m\": 4,\n \"mn\": \"Detailed protocol specification with state machines, legal grounding via FAA Section 2, and comprehensive dispute classification framework\",\n \"o\": 1,\n \"on\": \"Unique approach to agent transaction disputes - no existing IETF work addresses cryptographic attestation vs contractual satisfaction bifurcation\",\n \"mo\": 3,\n \"mon\": \"Active development with companion specifications co-submitted, but no clear WG adoption path or broader community engagement yet\",\n \"r\": 5,\n \"rn\": \"Core AI agent topic addressing fundamental challenge of autonomous transaction disputes with practical legal and technical framework\",\n \"c\": [\"A2A protocols\", \"Policy/governance\", \"Agent identity/auth\"]\n}", "in_tok": 957, "out_tok": 279} {"draft_name": "draft-hood-agtp-agent-cert", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft specifies an X.509 certificate-based identity verification extension for the Agent Transfer Protocol (AGTP), enabling infrastructure components to cryptographically verify agent identities and authority scopes at the transport layer. The extension includes session revocation mechanisms and Certificate Transparency logging for tamper-evident governance.\",\"n\":3,\"nn\":\"Useful contribution applying established PKI mechanisms to agent identity verification, with practical O(1) verification and governance features\",\"m\":4,\"mn\":\"Detailed specification with certificate extensions, revocation propagation, and CT logging mechanisms clearly defined\",\"o\":2,\"on\":\"While PKI and mutual TLS are well-established, the specific application to agent identity verification and AGTP integration shows minor similarities to existing work\",\"mo\":1,\"mon\":\"Future-dated draft (2026) with no visible development history or working group activity, appears to be initial proposal\",\"r\":5,\"rn\":\"Core AI agent infrastructure topic addressing fundamental identity and authentication challenges for autonomous agent communication\",\"c\":[\"Agent identity/auth\",\"A2A protocols\",\"Policy/governance\"]}", "in_tok": 725, "out_tok": 234} {"draft_name": "draft-han-ai-manifest", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft specifies a JSON-based AI Manifest protocol for websites to provide step-by-step UI workflow instructions to autonomous AI agents, enabling direct browser automation via CSS selectors. The specification includes three embedding methods, SHA-256 verification via a central trust registry, and reports an 81.9% reduction in input tokens with improved task success rates.\",\n \"n\": 4,\n \"nn\": \"Notable originality in creating a standardized manifest format specifically for AI agent browser automation, with the AFRM three-part model (frameworkHints, knownTraps, shortcuts) showing thoughtful design based on real-world practitioner needs.\",\n \"m\": 4,\n \"mn\": \"Detailed specification with concrete JSON schema, multiple embedding methods, verification procedures, and empirical performance data from reference implementation, though some security aspects could be more thoroughly specified.\",\n \"o\": 2,\n \"on\": \"Minor similarities to general web API documentation formats and robot.txt concepts, but the AI-specific workflow instruction approach and browser automation focus is largely unique.\",\n \"mo\": 3,\n \"mn\": \"Active development evident from multiple revisions (-01 update) with substantive improvements based on empirical findings, though no clear indication of working group adoption or broader community engagement.\",\n \"r\": 5,\n \"rn\": \"Core AI agent topic addressing fundamental challenges of web automation, token efficiency, and autonomous task execution - directly relevant to the future of AI agent interactions with web interfaces.\",\n \"c\": [\"Human-agent interaction\", \"Data formats/interop\", \"Agent discovery/reg\"]\n}", "in_tok": 909, "out_tok": 365} {"draft_name": "draft-li-atp", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"ATP defines a comprehensive communication protocol for autonomous agents operating across different deployment scales, featuring DNS-based discovery, mandatory authentication, and server-mediated routing. The protocol supports multiple interaction patterns including messaging, request/response, and event streaming for an Internet of Agents paradigm.\",\n \"n\": 4,\n \"nn\": \"Notable originality in defining a complete agent communication stack with scale-aware architecture and integrated discovery/auth mechanisms\",\n \"m\": 4,\n \"mn\": \"Detailed specification with comprehensive protocol definition, authentication framework, and message semantics across 57 pages\",\n \"o\": 2,\n \"on\": \"Minor similarities to existing messaging protocols but unique focus on autonomous agent communication at Internet scale\",\n \"mo\": 2,\n \"mon\": \"Single revision with future date suggests early proposal without established community engagement\",\n \"r\": 5,\n \"rn\": \"Core AI agent topic addressing fundamental infrastructure for autonomous agent communication and Internet of Agents\",\n \"c\": [\"A2A protocols\", \"Agent identity/auth\", \"Agent discovery/reg\"]\n}", "in_tok": 670, "out_tok": 252} {"draft_name": "draft-morrow-sogomonian-exec-outcome-attest", "model": "claude-sonnet-4-20250514", "raw": "```json\n{\n \"s\": \"This draft defines execution outcome verification as a new attestation primitive, introducing execution receipts to verify that AI agents and automated systems actually performed claimed actions with verifiable outcomes. It provides an abstract model and concrete mappings to transparency logs, TEE receipts, and other verification mechanisms.\",\n \"n\": 4,\n \"nn\": \"Notable originality in separating execution outcome verification from identity attestation as first-class concepts, with composable receipt primitives\",\n \"m\": 3,\n \"mn\": \"Defines protocol concepts and abstract model with concrete realizations, but likely needs more implementation details\",\n \"o\": 2,\n \"on\": \"Some conceptual overlap with general attestation frameworks but distinct focus on execution outcomes rather than identity\",\n \"mo\": 2,\n \"mon\": \"Single revision from April 2026, too early to assess broader momentum\",\n \"r\": 5,\n \"rn\": \"Core AI agent topic addressing fundamental trust and verification needs for autonomous system actions\",\n \"c\": [\"Agent identity/auth\", \"AI safety/alignment\"]\n}\n```", "in_tok": 597, "out_tok": 256} {"draft_name": "draft-mw-wimse-transitive-attestation", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This IETF draft defines a WIMSE profile for transitive attestation that cryptographically binds software workloads to their execution environment to prevent identity theft. It specifically addresses risks from AI agents moving across jurisdictions or being compromised through prompt injection attacks.\",\n \"n\": 4,\n \"nn\": \"Notable originality in addressing AI agent identity portability risks through transitive attestation chains - combines existing concepts in a novel way for emerging AI agent threats\",\n \"m\": 3,\n \"mn\": \"Defines a clear protocol mechanism with cryptographic binding approach, though lacks detailed implementation specifics and test vectors\",\n \"o\": 3,\n \"on\": \"Shares attestation concepts with RATS framework and references related WIMSE geofencing profile, but applies them distinctively to workload identity portability\",\n \"mo\": 4,\n \"mon\": \"Part of active WIMSE working group with clear references to companion drafts, indicating coordinated development effort\",\n \"r\": 5,\n \"rn\": \"Core AI agent topic - directly addresses autonomous AI agents, prompt injection attacks, and cross-jurisdictional movement challenges with concrete technical solutions\",\n \"c\": [\"Agent identity/auth\", \"AI safety/alignment\", \"Policy/governance\"]\n}", "in_tok": 808, "out_tok": 296} {"draft_name": "draft-oauth-transaction-tokens-for-agents", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft extends OAuth Transaction Tokens to support agent context propagation by using the 'act' field to identify the agent performing actions while the 'sub' field represents the principal. It enables more granular access control decisions for agent-based workloads within service call graphs.\",\"n\":2,\"nn\":\"Incremental extension applying existing OAuth transaction token patterns to agent scenarios\",\"m\":3,\"mn\":\"Defines clear protocol mechanism with field usage but lacks detailed examples and implementation guidance\",\"o\":4,\"on\":\"Significant overlap with base OAuth transaction tokens specification, primarily adding agent-specific field semantics\",\"mo\":3,\"mon\":\"Active development as evidenced by 2026 date, but momentum unclear without revision history\",\"r\":4,\"rn\":\"Directly addresses agent authentication and authorization, core concerns for AI agent deployments\",\"c\":[\"Agent identity/auth\"]}", "in_tok": 679, "out_tok": 189} {"draft_name": "draft-ietf-sml-trust", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft addresses trust and security considerations for structured email, providing recommendations for how email clients should handle structured data elements. It focuses on preventing abuse and ensuring secure processing of machine-readable email content.\",\"n\":2,\"nn\":\"Incremental security guidance for existing structured email concepts\",\"m\":2,\"mn\":\"Early sketch stage with basic recommendations, lacks detailed implementation guidance\",\"o\":3,\"on\":\"Shares security framework concepts with other email security drafts\",\"mo\":4,\"mon\":\"Active working group development with clear community engagement\",\"r\":2,\"rn\":\"Tangentially related - focuses on email security rather than AI agents specifically\",\"c\":[\"Data formats/interop\",\"Policy/governance\"]}", "in_tok": 579, "out_tok": 150} {"draft_name": "draft-cgfabk-nmrg-ibn-generative-ai", "model": "claude-sonnet-4-20250514", "raw": "```json\n{\n \"s\": \"This draft proposes methods for specializing generative AI models to create targeted solutions for Intent-Based Networking applications. It aims to provide a scalable and efficient approach to developing AI models that can understand and process network intents.\",\n \"n\": 3,\n \"nn\": \"Useful contribution combining two established fields (generative AI and IBN) in a practical way, though the core concept is somewhat predictable given current AI trends\",\n \"m\": 2,\n \"mn\": \"Early sketch stage - describes general approach for AI model specialization but lacks detailed protocols, specific mechanisms, or concrete implementation guidance\",\n \"o\": 3,\n \"on\": \"Shares conceptual overlap with existing IBN work and general AI networking drafts, but the specific focus on generative AI specialization is reasonably distinct\",\n \"mo\": 2,\n \"mon\": \"Single revision with future date (2026) suggests early/speculative work with limited development activity to date\",\n \"r\": 4,\n \"rn\": \"Directly relevant to AI agents in networking - addresses how to create specialized AI models for network intent processing and automation\",\n \"c\": [\"Autonomous netops\", \"Model serving/inference\", \"ML traffic mgmt\"]\n}\n```", "in_tok": 490, "out_tok": 293} {"draft_name": "draft-feng-nmrg-ain-architecture", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"Proposes an Internet-scale architecture called Agentic Intent Network (AIN) for enabling autonomous AI agents to discover and coordinate with each other across organizational boundaries without pre-established integrations. The draft presents architectural requirements and components for routing-based agent coordination but explicitly excludes engineering protocol details.\",\n \"n\": 4,\n \"nn\": \"Notable originality in applying Internet-scale routing principles to multi-agent coordination, addressing the fundamental challenge of agent discovery and invocation across organizational boundaries\",\n \"m\": 2,\n \"mn\": \"Early architectural sketch with high-level requirements and components defined, but lacks concrete protocol mechanisms, detailed specifications, or implementation guidance\",\n \"o\": 2,\n \"on\": \"Some conceptual similarities to service discovery and overlay networks, but the specific application to AI agent coordination represents a distinct approach\",\n \"mo\": 2,\n \"mon\": \"Appears to be initial submission with no indication of working group adoption or broader community engagement beyond NMRG research context\",\n \"r\": 5,\n \"rn\": \"Directly addresses core AI agent coordination challenges at Internet scale, representing a fundamental networking problem for the AI agent ecosystem\",\n \"c\": [\"A2A protocols\", \"Agent discovery/reg\", \"Data formats/interop\"]\n}", "in_tok": 619, "out_tok": 296} {"draft_name": "draft-borthwick-msebenzi-environment-state", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft defines the environment.* constraint family for Verifiable Intent mandates, enabling autonomous agents to check external conditions (venue status, fund availability, etc.) before executing transactions on behalf of humans. It provides the framework and vocabulary for environmental constraints but doesn't define specific constraint types.\",\n \"n\": 4,\n \"nn\": \"The concept of cryptographically binding environmental state constraints to autonomous agent mandates is notably original, bridging smart contract conditions with real-world state verification\",\n \"m\": 3,\n \"mn\": \"Defines a clear protocol framework with membership criteria, vocabulary, and composition rules, though lacks concrete constraint type definitions and implementation details\",\n \"o\": 2,\n \"on\": \"While smart contracts have conditional execution, this specific approach to environmental state verification in agent mandates appears relatively unique\",\n \"mo\": 2,\n \"mn\": \"Future-dated draft (2026) suggests early/conceptual stage with limited current activity or community engagement\",\n \"r\": 5,\n \"rn\": \"Directly addresses core autonomous agent functionality - how agents verify external conditions before acting on human behalf with cryptographic authorization\",\n \"c\": [\"A2A protocols\", \"Agent identity/auth\", \"Data formats/interop\", \"Human-agent interaction\"]\n}", "in_tok": 741, "out_tok": 295} {"draft_name": "draft-song-tsvwg-camp", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"CAMP proposes a consistency-aware multipath transport protocol built on MPQUIC to address latency and synchronization challenges in interactive multimodal LLM systems like digital humans. It introduces three-stream separation for different modalities, a transport-layer scheduler to reduce inter-modal arrival time deviations, and client-side alignment mechanisms.\",\n \"n\": 4,\n \"nn\": \"Novel combination of multipath transport with multimodal consistency guarantees specifically for LLM systems, introducing coordinated transport and application layer mechanisms\",\n \"m\": 3,\n \"mn\": \"Defines protocol architecture and key mechanisms but lacks detailed implementation specifications and test vectors for full deployment readiness\",\n \"o\": 2,\n \"on\": \"Builds on MPQUIC but addresses a distinct problem space of multimodal consistency with minimal overlap to existing transport work\",\n \"mo\": 2,\n \"mon\": \"Recent draft from emerging area but no clear indication of working group adoption or broader community engagement yet\",\n \"r\": 4,\n \"rn\": \"Directly addresses transport requirements for interactive AI systems, specifically targeting LLM-based applications with multimodal outputs\",\n \"c\": [\"ML traffic mgmt\", \"Model serving/inference\", \"Human-agent interaction\"]\n}", "in_tok": 776, "out_tok": 298} {"draft_name": "draft-yan-a2a-device-agent-applicability", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft explores applying the Agent-to-Agent (A2A) Protocol for communication between Controller Agents and Device Agents in network management, enabling network devices to become autonomous entities. It discusses workflows, deployment scenarios, and integration with existing protocols like NETCONF and RESTCONF.\",\n \"n\": 3,\n \"nn\": \"Useful application of A2A protocols to network management domain with specific focus on device-level agents\",\n \"m\": 2,\n \"mn\": \"Early sketch focusing on applicability without detailed protocol specifications or implementation details\",\n \"o\": 3,\n \"on\": \"Shares concepts with existing A2A protocol work and network management automation drafts\",\n \"mo\": 2,\n \"mon\": \"Single revision with future date suggests early-stage development without demonstrated community engagement\",\n \"r\": 5,\n \"rn\": \"Core AI agent topic focusing specifically on agent-to-agent communication protocols for autonomous network operations\",\n \"c\": [\"A2A protocols\", \"Autonomous netops\", \"Data formats/interop\"]\n}", "in_tok": 651, "out_tok": 252} {"draft_name": "draft-voit-rats-trustworthy-path-routing", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft proposes Trusted Path Routing to route sensitive traffic only through network devices that have been recently verified as trustworthy through attestation. It addresses scenarios where encryption alone is deemed insufficient and requires additional assurance about the integrity of transit infrastructure.\",\"n\":3,\"nn\":\"Useful contribution combining network routing with device attestation, though builds on existing RATS framework concepts\",\"m\":4,\"mn\":\"Detailed specification with concrete mechanisms for attestation-based routing decisions and path selection algorithms\",\"o\":3,\"on\":\"Shares attestation concepts with RATS working group drafts but applies them uniquely to routing decisions\",\"mo\":2,\"mon\":\"Single revision draft with limited visible community engagement or working group adoption\",\"r\":2,\"rn\":\"Tangentially related to AI through potential application to protecting AI traffic, but primarily a network security mechanism\",\"c\":[\"Other AI/agent\"]}", "in_tok": 562, "out_tok": 191} {"draft_name": "draft-howard-virp", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"VIRP defines a cryptographic trust framework for verifying operations performed by human or autonomous agents on live network infrastructure. It provides seven trust primitives including observation integrity, intent gating, and outcome verification with formal protocol analysis and multi-vendor implementation results.\",\"n\":4,\"nn\":\"Novel cryptographic dual-signature scheme and trust tier architecture specifically designed for autonomous network operations\",\"m\":5,\"mn\":\"Complete specification with live multi-vendor implementation, formal verification in ProVerif/Tamarin, and detailed test results\",\"o\":2,\"on\":\"Some conceptual overlap with general infrastructure security protocols but unique focus on autonomous agent operations\",\"mo\":2,\"mon\":\"Single draft with no clear working group adoption or community discussion visible\",\"r\":5,\"rn\":\"Directly addresses core challenge of trusting autonomous agents operating on critical network infrastructure\",\"c\":[\"Autonomous netops\",\"Agent identity/auth\",\"AI safety/alignment\"]}", "in_tok": 896, "out_tok": 198} {"draft_name": "draft-xiong-rtgwg-use-cases-hp-wan", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft identifies use cases for high-performance wide area networks driven by big data and AI computing demands. It appears to be a problem statement document cataloging network performance requirements for emerging AI/ML applications.\",\"n\":1,\"nn\":\"Basic enumeration of known requirements for AI/ML network performance - no novel technical insights\",\"m\":2,\"mn\":\"Early stage use case collection without detailed technical specifications or solutions\",\"o\":4,\"on\":\"Significant overlap with existing network optimization and SDN work, plus general AI networking requirements\",\"mo\":2,\"mon\":\"Single revision draft with no clear working group adoption or community engagement\",\"r\":3,\"rn\":\"Partially relevant as it addresses networking for AI/ML workloads but focuses on infrastructure rather than agent protocols\",\"c\":[\"ML traffic mgmt\"]}", "in_tok": 512, "out_tok": 172} {"draft_name": "draft-chang-delta1-settlement", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"Delta-1 defines a SCITT profile for generating binary accountability receipts that prove AI agent decisions met three specific conditions: evidence sealing, strategic intent isolation, and authorized settlement recording. The profile targets regulated industries needing verifiable AI decision audit trails with offline verification capabilities.\",\n \"n\": 4,\n \"nn\": \"Notable application of SCITT framework to AI accountability with specific three-condition model and binary settlement approach - creative adaptation of supply chain integrity concepts to AI decision verification\",\n \"m\": 3,\n \"mn\": \"Defines clear protocol mechanics with three accountability conditions and binary verdict model, though lacks implementation details and test vectors for production deployment\",\n \"o\": 2,\n \"on\": \"Builds on SCITT framework but applies it uniquely to AI decision accountability - minimal overlap with existing AI agent specifications in IETF\",\n \"mo\": 2,\n \"mon\": \"Single draft publication with no evident working group adoption or community discussion - appears to be individual contribution without broader momentum\",\n \"r\": 4,\n \"rn\": \"Directly addresses AI agent accountability and decision verification, which are core concerns for agentic AI deployment in regulated environments\",\n \"c\": [\"Policy/governance\", \"Agent identity/auth\", \"Data formats/interop\"]\n}", "in_tok": 705, "out_tok": 297} {"draft_name": "draft-zheng-agent-identity-management", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft specifies agent identity management for Internet of Agents (IOA) systems, defining requirements for agent identities, registration processes, identifier structures, and identity management functions performed by agent gateways. It appears to be an early framework for managing autonomous agent identities in networked environments.\",\"n\":3,\"nn\":\"Useful contribution addressing agent identity in networked systems, though identity management concepts are well-established in other domains\",\"m\":2,\"mn\":\"Early sketch with basic framework defined but lacks detailed protocol specifications, implementation details, or concrete examples\",\"o\":3,\"on\":\"Shares general identity management concepts with existing work but applies them specifically to agent systems\",\"mo\":1,\"mon\":\"Single draft submission with no evidence of working group activity or community engagement\",\"r\":5,\"rn\":\"Core AI agent topic directly addressing fundamental identity and authentication needs for autonomous agent systems\",\"c\":[\"Agent identity/auth\",\"Agent discovery/reg\",\"A2A protocols\"]}", "in_tok": 519, "out_tok": 206} {"draft_name": "draft-rosenberg-aiproto-nact", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"N-ACT defines a standardized protocol for AI agents to discover and invoke third-party tools through enumeration and invocation APIs. The goal is to reduce integration costs for enterprise AI agents needing to use external APIs.\",\n \"n\": 2,\n \"nn\": \"Incremental work applying existing API design patterns to AI agent tool integration - straightforward but useful standardization effort\",\n \"m\": 3,\n \"mn\": \"Defines core protocol components (enumeration and invocation APIs) with clear problem statement, though lacks detailed implementation specifics\",\n \"o\": 3,\n \"on\": \"Shares conceptual ground with OpenAI function calling specs and general API gateway patterns, but focuses specifically on AI agent use cases\",\n \"mo\": 2,\n \"mon\": \"Single draft submission with no indication of working group adoption or broader community involvement yet\",\n \"r\": 4,\n \"rn\": \"Directly addresses a practical need in AI agent deployments - tool integration is a core challenge for enterprise AI systems\",\n \"c\": [\"A2A protocols\", \"Data formats/interop\"]\n}", "in_tok": 722, "out_tok": 261} {"draft_name": "draft-li-sidrops-drr-architecture", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"Proposes a decentralized RPKI repository architecture (dRR) that separates signing authority from object management through distributed certificate servers and monitors. Addresses current RPKI limitations around single points of failure, scalability, and consistency while maintaining compatibility with existing trust models.\",\n \"n\": 3,\n \"nn\": \"Useful contribution applying decentralization concepts to RPKI infrastructure challenges, though builds on established distributed systems principles\",\n \"m\": 3,\n \"mn\": \"Defines the architectural components and approach but appears to lack detailed protocol specifications and implementation guidance\",\n \"o\": 2,\n \"on\": \"Minor similarities to other distributed trust infrastructure proposals but addresses specific RPKI repository concerns\",\n \"mo\": 2,\n \"mon\": \"Recent draft with limited revision history, unclear working group status or community engagement\",\n \"r\": 1,\n \"rn\": \"Focuses on RPKI infrastructure security and reliability, not related to AI or autonomous agents\",\n \"c\": [\"Other AI/agent\"]\n}", "in_tok": 760, "out_tok": 244} {"draft_name": "draft-yan-iba-routing-security-requirements", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft specifies security requirements for intent-based agent routing, defining a security architecture with phase-specific attack analysis for Registration, Resolution, and Dispatch phases. It addresses new security risks beyond traditional channel protection when autonomous agents collaborate based on semantic intent rather than static addresses.\",\"n\":4,\"nn\":\"Notable originality in addressing semantic intent security risks that go beyond traditional transport-layer protections\",\"m\":3,\"mn\":\"Defines security architecture and requirements framework but appears to be at requirements specification rather than detailed implementation level\",\"o\":2,\"on\":\"Minor overlap with general agent security work but unique focus on intent-based routing security\",\"mo\":2,\"mon\":\"Single revision with April 2026 date suggests early stage work\",\"r\":5,\"rn\":\"Core AI agent topic addressing fundamental security challenges in autonomous agent collaboration and routing\",\"c\":[\"A2A protocols\",\"AI safety/alignment\",\"Agent identity/auth\",\"Agent discovery/reg\"]}", "in_tok": 665, "out_tok": 202} {"draft_name": "draft-han-ai-agent-impact-infra", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft analyzes how AI agents impact network infrastructure to support IETF standardization discussions. It appears to be an early analysis document examining the networking implications of AI agent deployment and communication.\",\"n\":2,\"nn\":\"Incremental analysis of known AI agent networking challenges\",\"m\":1,\"mn\":\"Problem statement level with basic analysis, lacks concrete protocols or mechanisms\",\"o\":3,\"on\":\"Shares general concepts with other AI networking drafts but focuses specifically on infrastructure impact\",\"mo\":2,\"mon\":\"Single early draft with basic scope, unclear community engagement\",\"r\":4,\"rn\":\"Directly addresses AI agent networking infrastructure needs, highly relevant to IETF AI standardization\",\"c\":[\"Autonomous netops\",\"A2A protocols\",\"ML traffic mgmt\"]}", "in_tok": 484, "out_tok": 168} {"draft_name": "draft-lkspa-wimse-verifiable-geo-fence", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft proposes a hardware-rooted cryptographic framework for verifying workload location and integrity using zero-knowledge proofs to ensure privacy. It establishes a 'Silicon-to-Workload' chain of trust that binds software identities to persistent hardware identities and verified geographic residency.\",\n \"n\": 4,\n \"nn\": \"Notable combination of hardware attestation, zero-knowledge proofs, and geographic verification for workload identity - creative approach to sovereignty requirements\",\n \"m\": 2,\n \"mn\": \"Early conceptual stage with high-level architecture described but lacking detailed protocol specifications, message formats, or implementation guidance\",\n \"o\": 2,\n \"on\": \"Some conceptual overlap with hardware attestation and identity work, but the privacy-preserving geofencing approach appears relatively unique\",\n \"mo\": 2,\n \"mon\": \"Single draft submission with no clear working group adoption or community feedback visible\",\n \"r\": 4,\n \"rn\": \"Directly addresses workload identity management for AI/ML systems with focus on sovereign computing requirements and hardware-backed trust\",\n \"c\": [\"Agent identity/auth\", \"Policy/governance\"]\n}", "in_tok": 835, "out_tok": 277} {"draft_name": "draft-aevum-agentcard", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"AgentCard defines a JSON format for autonomous software agents to declare their identity, capabilities, and communication details in a framework-neutral way. It enables interoperability between agents built with different frameworks like LangChain, CrewAI, or AutoGen without prior coordination.\",\n \"n\": 4,\n \"nn\": \"Notable originality in creating a framework-neutral identity format with physics-grounded energy pricing and ULID-based global identity\",\n \"m\": 4,\n \"mn\": \"Detailed specification with comprehensive schema elements, pricing mechanisms, and clear integration patterns\",\n \"o\": 2,\n \"on\": \"Minor similarities to general service discovery formats but unique approach to agent-specific identity and capability declaration\",\n \"mo\": 2,\n \"mon\": \"Single revision from 2026 with no apparent working group adoption or community momentum indicators\",\n \"r\": 5,\n \"rn\": \"Core AI agent topic addressing fundamental interoperability challenges in multi-agent systems\",\n \"c\": [\"A2A protocols\", \"Agent identity/auth\", \"Data formats/interop\", \"Agent discovery/reg\"]\n}", "in_tok": 702, "out_tok": 266} {"draft_name": "draft-farzdusa-aipref-enduser", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft examines the potential impact on end users of standardizing AI preference signaling mechanisms, focusing on technological, legal, ethical, and governance dimensions. It aims to provide principles for the ai-pref working group to consider when developing vocabulary and definitions for AI preference signaling standards.\",\n \"n\": 2,\n \"nn\": \"Incremental contribution focusing on end-user impact analysis of AI preference signaling, which is a useful but not particularly novel perspective on an emerging standardization area\",\n \"m\": 2,\n \"mn\": \"Early sketch providing impact analysis and suggested principles but lacking concrete protocol mechanisms, detailed specifications, or implementation guidance\",\n \"o\": 3,\n \"on\": \"Shares conceptual space with other AI preference and signaling drafts, particularly around user control and data privacy, but focuses specifically on end-user impact analysis\",\n \"mo\": 3,\n \"mon\": \"Recent draft (2025-11-27) addressing an active area of IETF interest with the ai-pref working group, showing ongoing development in AI preference signaling standardization\",\n \"r\": 4,\n \"rn\": \"Directly relevant to AI agent systems as it addresses how users interact with and control AI systems through preference signaling mechanisms, focusing on user agency and control\",\n \"c\": [\"Policy/governance\", \"Human-agent interaction\", \"Data formats/interop\"]\n}", "in_tok": 559, "out_tok": 322} {"draft_name": "draft-zeng-nmrg-mcp-usecases-requirements", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft proposes integrating the Model Context Protocol (MCP) into network management to enable AI agents to interact with network infrastructure through standardized, secure interfaces. It presents problem statements, use cases for troubleshooting and optimization, and requirements for MCP-based network management architectures.\",\"n\":3,\"nn\":\"Useful contribution applying MCP to network management domain, though MCP itself is established\",\"m\":2,\"mn\":\"Early stage problem statement and requirements document without detailed protocol specifications\",\"o\":2,\"on\":\"Minor conceptual overlap with general network automation but unique focus on MCP integration\",\"mo\":2,\"mon\":\"Single revision draft with no clear working group adoption signals\",\"r\":4,\"rn\":\"Directly relevant to AI agents in networking with concrete use cases and architectural considerations\",\"c\":[\"Autonomous netops\",\"A2A protocols\",\"Data formats/interop\"]}", "in_tok": 595, "out_tok": 191} {"draft_name": "draft-lake-pocero-authkem-edhoc", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft extends EDHOC protocol with post-quantum cryptography mechanisms, specifically adding KEM-based authentication to provide quantum resistance. It defines signature-free authentication methods using standardized post-quantum KEMs like ML-KEM.\",\"n\":3,\"nn\":\"Useful contribution applying established PQC concepts to EDHOC authentication, though the core idea of KEM-based auth is known\",\"m\":4,\"mn\":\"Detailed specification with concrete protocol extensions, message formats, and integration with existing EDHOC framework\",\"o\":3,\"on\":\"Shares PQC integration concepts with other quantum-resistant protocol extensions but specific to EDHOC authentication\",\"mo\":3,\"mon\":\"Active development with specific focus on standardizing PQC in authentication protocols\",\"r\":2,\"rn\":\"Security protocol enhancement rather than AI/agent-focused, only tangentially related through potential use in agent authentication\",\"c\":[\"Agent identity/auth\"]}", "in_tok": 570, "out_tok": 206} {"draft_name": "draft-han-rtgwg-agent-gateway-intercomm-framework", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft defines a framework for intercommunication between Agent Gateways in an Agent Internet ecosystem, specifying hierarchical layered models and protocol requirements. It aims to enable cross-domain agent collaboration through standardized gateway interoperability mechanisms.\",\"n\":3,\"nn\":\"Useful contribution to agent internetworking - builds on IoA concepts but focuses specifically on gateway intercommunication patterns\",\"m\":2,\"mn\":\"Early framework sketch - defines high-level architecture and requirements but lacks detailed protocol specifications or implementation guidance\",\"o\":4,\"on\":\"Significant conceptual overlap with existing IoA reference architectures and general gateway interoperability frameworks\",\"mo\":2,\"mon\":\"Single revision with no evidence of working group adoption or broader community engagement\",\"r\":4,\"rn\":\"Directly relevant to AI agent networking - addresses core challenges of cross-domain agent communication and collaboration\",\"c\":[\"A2A protocols\",\"Data formats/interop\",\"Agent discovery/reg\"]}", "in_tok": 582, "out_tok": 206} {"draft_name": "draft-ietf-dtn-dtnma", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft defines a management architecture for Delay-Tolerant Networks (DTN) that operates asynchronously without query-response mechanisms, designed for challenged environments with long delays and frequent disruptions. The architecture supports autonomous local control with minimal resource footprint for constrained devices operating over the Bundle Protocol.\",\"n\":3,\"nn\":\"Useful contribution adapting network management principles to DTN's unique constraints, though builds on established DTN and network management concepts\",\"m\":4,\"mn\":\"Detailed 59-page specification with comprehensive architecture definition, appears implementation-ready with thorough treatment of DTN management challenges\",\"o\":2,\"on\":\"Some conceptual overlap with general network management architectures but DTN-specific approach is fairly unique\",\"mo\":4,\"mon\":\"IETF working group draft suggesting active development and community interest in DTN management standardization\",\"r\":3,\"rn\":\"Partially relevant due to autonomous local control aspects and asynchronous operation, though primarily focused on network management rather than AI agents\",\"c\":[\"Autonomous netops\"]}", "in_tok": 661, "out_tok": 227} {"draft_name": "draft-jeskey-anml", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"ANML defines a machine-first markup language for agent-to-agent communication, providing structured formats for content, interactions, constraints, and behavioral guidance. The 80-page specification offers both XML and JSON serializations to enable deterministic agent interactions with reduced reliance on unstructured data inference.\",\n \"n\": 4,\n \"nn\": \"Novel approach to standardizing agent communication with comprehensive framework including persona, constraints, and state management - goes beyond simple API definitions\",\n \"m\": 4,\n \"mn\": \"Detailed 80-page specification with dual serialization formats, abstract data model, and comprehensive feature set suggests implementation-ready maturity\",\n \"o\": 2,\n \"on\": \"While agent communication exists, the holistic approach combining content, interactions, knowledge, constraints, state, and persona appears relatively unique\",\n \"mo\": 3,\n \"mon\": \"Recent 2026 date suggests active development, though single draft snapshot limits momentum assessment\",\n \"r\": 5,\n \"rn\": \"Core AI agent topic addressing fundamental challenge of standardized agent-to-agent communication protocols\",\n \"c\": [\"A2A protocols\", \"Data formats/interop\", \"Agent identity/auth\", \"Policy/governance\"]\n}", "in_tok": 770, "out_tok": 289} {"draft_name": "draft-chen-oauth-scope-agent-extensions", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft extends OAuth 2.0 scopes with structured syntax to enable fine-grained authorization for AI agent capabilities and modular functional units. It proposes a colon-separated format [resource_type]:[action]:[target][:constraints] to standardize permission requests for agent operations like file access, network access, and tool invocation.\",\n \"n\": 4,\n \"nn\": \"Notable originality in applying structured OAuth scopes specifically to agent capability authorization - addresses a real gap in current agent security models\",\n \"m\": 3,\n \"mn\": \"Defines clear protocol extension with syntax specification, though lacks detailed examples and implementation guidance\",\n \"o\": 2,\n \"on\": \"Minor similarities to general OAuth extension work, but the agent-specific capability authorization approach is relatively unique\",\n \"mo\": 2,\n \"mon\": \"Appears to be early stage work without clear working group adoption or community feedback\",\n \"r\": 5,\n \"rn\": \"Directly addresses core AI agent security concerns around capability authorization and permission management\",\n \"c\": [\"Agent identity/auth\", \"Policy/governance\", \"AI safety/alignment\"]\n}", "in_tok": 739, "out_tok": 272} {"draft_name": "draft-zhang-rtgwg-agent-policy-aware-network", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft proposes an 'AI Agent Policy-Aware Network' paradigm that enables network infrastructure to understand and respond to AI agent intents rather than just traffic flows. It defines components like Agent Policy-aware Controllers to create dynamic mappings between agent policies and network configurations for scenarios like autonomous performance measurement and path optimization.\",\n \"n\": 3,\n \"nn\": \"Useful contribution applying intent-based networking concepts to AI agents, though builds on established policy-aware networking principles\",\n \"m\": 2,\n \"mn\": \"Early sketch defining high-level components and use cases but lacking detailed protocols, mechanisms, or concrete specifications\",\n \"o\": 3,\n \"on\": \"Shares concepts with intent-based networking and policy-aware systems, but specific application to AI agents is distinct\",\n \"mo\": 2,\n \"mon\": \"Single revision draft with no clear working group adoption or community discussion yet\",\n \"r\": 4,\n \"rn\": \"Directly relevant to AI agents in networking, addressing how network infrastructure can become agent-aware and policy-driven\",\n \"c\": [\"Autonomous netops\", \"Policy/governance\", \"ML traffic mgmt\"]\n}", "in_tok": 643, "out_tok": 275} {"draft_name": "draft-song-dmsc-problem-statement", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft identifies problems with current LLM-based multi-agent systems where each agent must independently handle communication, security, and collaboration logic, leading to inefficiency and security risks. It proposes requirements for a centralized DMSC infrastructure to offload these common functions and enable better scalability in multi-tenant environments.\",\n \"n\": 3,\n \"nn\": \"Centralizing multi-agent infrastructure functions is a logical evolution but represents a useful architectural contribution to the space\",\n \"m\": 1,\n \"mn\": \"Pure problem statement document without defined protocols, mechanisms, or technical specifications\",\n \"o\": 3,\n \"on\": \"Shares architectural concepts with service mesh and API gateway approaches, though applied specifically to AI agents\",\n \"mo\": 2,\n \"mon\": \"Single draft from 2026 with no apparent working group adoption or community engagement yet\",\n \"r\": 5,\n \"rn\": \"Directly addresses core challenges in multi-agent AI systems including communication, security, and collaboration coordination\",\n \"c\": [\"A2A protocols\", \"Agent identity/auth\", \"Policy/governance\", \"Agent discovery/reg\"]\n}", "in_tok": 664, "out_tok": 270} {"draft_name": "draft-cui-dns-native-agent-naming-resolution", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft proposes DNS-Native Agent Naming and Resolution (DN-ANR) to use FQDNs as stable identifiers for AI agents and resolve them to verifiable endpoints with cryptographic integrity. It focuses on minimal pointer capabilities while avoiding heavy semantic metadata in DNS.\",\n \"n\": 3,\n \"nn\": \"Useful contribution applying established DNS patterns to AI agent identification - not groundbreaking but addresses a clear need\",\n \"m\": 4,\n \"mn\": \"Detailed specification with clear protocol mechanisms and examples, appears implementation-ready\",\n \"o\": 2,\n \"on\": \"Minor similarities to general service discovery mechanisms but specific application to AI agents is relatively unique\",\n \"mo\": 2,\n \"mon\": \"Single revision draft with no clear working group adoption or broader community engagement yet\",\n \"r\": 4,\n \"rn\": \"Directly addresses AI agent infrastructure needs for naming, discovery and verification - core to agent ecosystem development\",\n \"c\": [\"Agent identity/auth\", \"Agent discovery/reg\", \"A2A protocols\"]\n}", "in_tok": 599, "out_tok": 253} {"draft_name": "draft-ietf-drip-auth", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft defines authentication message formats and protocols for the Drone Remote Identification Protocol (DRIP) to enable real-time verification of drone broadcast messages. It specifies cryptographic mechanisms to authenticate that Remote ID messages were signed by registered drone operators, supporting offline verification by observers without Internet access.\",\"n\":2,\"nn\":\"Incremental extension applying standard cryptographic authentication to drone identification - useful but builds on established patterns\",\"m\":4,\"mn\":\"Detailed 48-page specification with defined message formats, authentication protocols, and specific implementation guidance\",\"o\":3,\"on\":\"Shares authentication and identity verification concepts with other Remote ID and aircraft identification standards\",\"mo\":4,\"mon\":\"Active IETF working group draft with recent updates, indicating sustained development effort\",\"r\":3,\"rn\":\"Partially relevant as it addresses autonomous system identity and authentication, though focused on drones rather than AI agents broadly\",\"c\":[\"Agent identity/auth\",\"Policy/governance\"]}", "in_tok": 585, "out_tok": 207} {"draft_name": "draft-dunbar-agent-attachment", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"Defines Agent Attachment Protocol (AAP) for AI agents to establish connections to edge nodes and obtain communication properties like identifiers and attachment context. Focuses on maintaining attachment state and representing properties for use by forwarding functions and routing mechanisms.\",\n \"n\": 3,\n \"nn\": \"Useful contribution addressing the specific problem of agent-to-infrastructure attachment, though builds on familiar networking attachment concepts\",\n \"m\": 3,\n \"mn\": \"Appears to define a complete protocol mechanism with attachment procedures and property representation, though lacks implementation details\",\n \"o\": 3,\n \"on\": \"Shares networking attachment concepts with existing protocols but applies them specifically to AI agents in edge computing contexts\",\n \"mo\": 2,\n \"mon\": \"Single draft from individual author with no clear working group activity or community engagement indicated\",\n \"r\": 4,\n \"rn\": \"Directly addresses AI agent networking needs with focus on attachment protocols and communication mechanisms for agent infrastructure\",\n \"c\": [\"A2A protocols\", \"Agent discovery/reg\", \"Agent identity/auth\"]\n}", "in_tok": 570, "out_tok": 253} {"draft_name": "draft-chu-oauth-as-attested-user-cert", "model": "claude-sonnet-4-20250514", "raw": "```json\n{\n \"s\": \"Extends OAuth 2.0 Rich Authorization Requests to allow clients (like AI agents) to request AS-attested user certificates within access tokens. Enables resource servers to obtain trusted user public keys for verifying delegation evidence signed by users.\",\n \"n\": 3,\n \"nn\": \"Useful contribution combining OAuth RAR with certificate attestation for delegation scenarios, particularly relevant for AI agents needing verified user authority\",\n \"m\": 2,\n \"mn\": \"Early sketch stage - introduces concept and basic mechanism but lacks detailed protocol specification, examples, and implementation details\",\n \"o\": 3,\n \"on\": \"Shares concepts with OAuth extensions and certificate-based authentication drafts, but specific combination for AI agent delegation appears less common\",\n \"mo\": 2,\n \"mon\": \"Single revision draft with 2026 date suggests recent submission but no visible working group adoption or community engagement yet\",\n \"r\": 4,\n \"rn\": \"Directly addresses AI agent authentication and authorization needs, specifically tackling how agents can obtain verifiable user delegation authority\",\n \"c\": [\"Agent identity/auth\", \"A2A protocols\"]\n}\n```", "in_tok": 595, "out_tok": 274} {"draft_name": "draft-sharif-openid-agent-identity", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft defines an OpenID Connect profile for issuing identity tokens to autonomous AI agents, introducing standard claims for agent identity, ownership, capabilities, and trust levels. It enables existing OIDC infrastructure to authenticate and authorize AI agents without core protocol modifications.\",\n \"n\": 3,\n \"nn\": \"Useful adaptation of existing identity standards to AI agents, though the core concept of extending OIDC for non-human entities is not entirely novel\",\n \"m\": 2,\n \"mn\": \"At 2 pages, this appears to be an early sketch outlining the concept rather than a detailed specification with implementation guidance\",\n \"o\": 2,\n \"on\": \"Some conceptual overlap with general identity federation work, but the specific focus on AI agent identity claims appears relatively unique\",\n \"mo\": 2,\n \"mon\": \"Single draft from 2026 with no indication of working group adoption or broader community engagement\",\n \"r\": 5,\n \"rn\": \"Directly addresses a core challenge in AI agent deployment - establishing trusted identity and authentication mechanisms for autonomous systems\",\n \"c\": [\"Agent identity/auth\", \"Policy/governance\"]\n}", "in_tok": 587, "out_tok": 271} {"draft_name": "draft-chen-tsvwg-high-speed-data-express", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"Proposes a high-speed data express architecture for IP networks to support massive data transmission for AI and supercomputing applications. Introduces a 'one low and three high' technical system for differentiated data express services.\",\n \"n\": 2,\n \"nn\": \"Incremental approach applying existing IP network optimization concepts to AI data transmission needs\",\n \"m\": 2,\n \"mn\": \"Early sketch with high-level architecture and concepts but lacks detailed protocol specifications or implementation details\",\n \"o\": 3,\n \"on\": \"Shares general concepts with existing high-speed networking and QoS drafts, though specific focus on AI data express is somewhat distinct\",\n \"mo\": 1,\n \"mon\": \"Appears to be initial draft with no visible community engagement or working group activity\",\n \"r\": 3,\n \"rn\": \"Partially relevant as it addresses infrastructure needs for AI applications, though not directly about AI agents themselves\",\n \"c\": [\"ML traffic mgmt\", \"Other AI/agent\"]\n}", "in_tok": 576, "out_tok": 244} {"draft_name": "draft-reilly-rmrp", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"RMRP defines a protocol for routing AI inference requests across multiple models with policy governance and audit trails. It provides vendor-agnostic framework for managing routing decisions, cost attribution, and compliance in multi-model AI deployments.\",\n \"n\": 4,\n \"nn\": \"First comprehensive protocol specifically addressing AI model routing with policy governance - novel intersection of network routing concepts with AI infrastructure management\",\n \"m\": 4,\n \"mn\": \"At 37 pages with detailed metadata structures, routing policies, execution semantics, and audit requirements - appears to be a well-developed specification\",\n \"o\": 2,\n \"on\": \"Some conceptual overlap with general routing protocols and service mesh concepts, but AI model routing with policy governance is largely unexplored territory\",\n \"mo\": 2,\n \"mon\": \"Future-dated draft (2026) suggests speculative work without current community development or validation\",\n \"r\": 5,\n \"rn\": \"Directly addresses core AI infrastructure challenge of managing requests across multiple models with governance requirements - highly relevant to AI agent ecosystems\",\n \"c\": [\"ML traffic mgmt\", \"Policy/governance\", \"Model serving/inference\", \"Data formats/interop\"]\n}", "in_tok": 638, "out_tok": 284} {"draft_name": "draft-beyer-agent-identity-architecture", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"Defines an architectural model for human-anchored agent identity that establishes delegation semantics and provenance structures to ensure agents can be traced back to responsible humans. The framework addresses the critical problem of accountability in autonomous agent systems without defining specific protocols.\",\n \"n\": 4,\n \"nn\": \"Notable originality in proposing human identity roots and explicit delegation chains for agent accountability - addresses a gap not well covered by existing identity frameworks\",\n \"m\": 2,\n \"mn\": \"Early architectural sketch that defines concepts and models but lacks concrete protocols, wire formats, or implementation details\",\n \"o\": 2,\n \"on\": \"Minor similarities to general identity and delegation frameworks, but the human-anchored agent focus is relatively unique in IETF space\",\n \"mo\": 2,\n \"mon\": \"Single revision with future date suggests early stage work - unclear if there's working group interest or broader community engagement\",\n \"r\": 5,\n \"rn\": \"Core AI agent topic addressing fundamental challenges of accountability, delegation, and provenance in autonomous agent ecosystems\",\n \"c\": [\"Agent identity/auth\", \"AI safety/alignment\", \"Human-agent interaction\", \"Policy/governance\"]\n}", "in_tok": 653, "out_tok": 279} {"draft_name": "draft-beyer-agent-identity-avian-carriers", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This April 1st RFC draft humorously proposes cryptographic identity and provenance attestation for AI agents communicating via carrier pigeons, extending the classic RFC 1149 IP over Avian Carriers joke. It addresses delegation chains and hallucination propagation across 'feather-based transport layers' with tongue-in-cheek technical rigor.\",\n \"n\": 5,\n \"nn\": \"Genuinely creative fusion of AI agent identity concepts with the classic avian carrier RFCs - unprecedented technical humor territory\",\n \"m\": 2,\n \"mn\": \"Early conceptual sketch with humor taking precedence over technical depth, lacks implementation details beyond the joke framework\",\n \"o\": 1,\n \"on\": \"Completely unique approach combining agent identity with avian carriers - no other work remotely similar\",\n \"mo\": 1,\n \"mn\": \"April 1st publication with no real development momentum expected or intended\",\n \"r\": 3,\n \"rn\": \"Partially relevant as it does address real AI agent concepts like delegation chains and provenance, albeit in humorous context\",\n \"c\": [\"Agent identity/auth\", \"Other AI/agent\"]\n}", "in_tok": 624, "out_tok": 281} {"draft_name": "draft-romm-aipref-contentsignals", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft proposes a vocabulary for three categories of content signals that enable parties to express preferences about how their digital assets are used by automated processing systems. It extends the AIPREF-VOCAB framework to provide granular control over AI system interactions with content.\",\"n\":2,\"nn\":\"Incremental extension of existing AIPREF vocabulary work, categorizing content signals in a straightforward manner\",\"m\":2,\"mn\":\"Early sketch level - only 5 pages describing categories without detailed mechanisms, examples, or implementation guidance\",\"o\":4,\"on\":\"Significant overlap with existing AIPREF-VOCAB work and other content preference/rights expression efforts\",\"mo\":2,\"mon\":\"Single revision with October 2025 date suggests minimal development activity so far\",\"r\":4,\"rn\":\"Directly relevant to AI systems as it addresses how content owners can signal usage preferences to automated processing systems\",\"c\":[\"Policy/governance\",\"Data formats/interop\"]}", "in_tok": 513, "out_tok": 208} {"draft_name": "draft-jain-lisp-network-ai-infra", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"This draft proposes using LISP (Locator/Identifier Separation Protocol) as a unified control plane for AI infrastructure networks, enabling scale-up, scale-out, and scale-across deployment patterns. It leverages existing LISP mechanisms like EID-to-RLOC mapping and VPN segmentation to provide scalable bandwidth and low-latency operations for AI training and inference workloads across heterogeneous accelerator fabrics.\",\"n\":2,\"nn\":\"Incremental application of existing LISP protocol to AI infrastructure use case\",\"m\":3,\"mn\":\"Defines mechanisms and architecture but lacks detailed implementation specifics\",\"o\":3,\"on\":\"Shares networking concepts with other AI infrastructure drafts but takes distinct LISP-based approach\",\"mo\":2,\"mon\":\"Single revision, unclear WG interest or community engagement\",\"r\":3,\"rn\":\"Partially relevant - focuses on network infrastructure for AI rather than AI agents themselves\",\"c\":[\"ML traffic mgmt\"]}", "in_tok": 710, "out_tok": 219} {"draft_name": "draft-li-rtgwg-distributed-lossless-framework", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"Proposes a framework for distributed AI data center networks to enable lossless interconnection between multiple data centers for large-scale model training. Focuses on networking challenges for connecting geographically distributed data centers into unified AI training clusters.\",\"n\":2,\"nn\":\"Incremental approach applying existing lossless networking concepts to AI data center interconnection without significant technical innovation\",\"m\":2,\"mn\":\"Early framework sketch with high-level concepts but lacks detailed protocols, mechanisms, or implementation specifics\",\"o\":3,\"on\":\"Shares networking concepts with existing data center interconnect work but applies them specifically to AI workloads\",\"mo\":2,\"mon\":\"Single revision draft with no evidence of working group adoption or broader community engagement\",\"r\":4,\"rn\":\"Directly addresses AI infrastructure networking needs for distributed training, which is highly relevant to AI systems\",\"c\":[\"ML traffic mgmt\",\"Other AI/agent\"]}", "in_tok": 581, "out_tok": 197} {"draft_name": "draft-men-rtgwg-agent-networking-digibank-scenario", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft proposes an agent networking architecture centered on agent gateways for digital banking scenarios, discussing how banking digitalization is evolving toward agentic service interconnection. It appears to be an early-stage exploration of applying agent networking concepts to enterprise banking use cases.\",\n \"n\": 2,\n \"nn\": \"Incremental application of agent networking concepts to banking domain, but architecture appears to be straightforward gateway-based approach without novel technical contributions\",\n \"m\": 2,\n \"mn\": \"Early sketch level - describes scenarios and proposes basic architecture but lacks detailed protocol specifications, implementation details, or concrete mechanisms\",\n \"o\": 3,\n \"on\": \"Shares general agent networking and gateway concepts with other drafts in the space, though banking-specific application provides some differentiation\",\n \"mo\": 2,\n \"mon\": \"Single revision draft with no clear working group adoption or broader community engagement visible\",\n \"r\": 4,\n \"rn\": \"Directly relevant to AI agent networking, focusing on agent interconnection architecture and service communication in enterprise environments\",\n \"c\": [\"A2A protocols\", \"Agent discovery/reg\", \"Data formats/interop\"]\n}", "in_tok": 515, "out_tok": 276} {"draft_name": "draft-yao-tsvwg-cco-requirement-and-analysis", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft analyzes network-assisted collective communication optimizations for AI training/inference clusters, identifying gaps between emerging RDMA-based acceleration approaches and current IETF standards like iWARP. It proposes requirements for new standards to optimize Many-to-Many communication patterns essential for large-scale AI workloads.\",\n \"n\": 3,\n \"nn\": \"Useful contribution identifying specific gaps between AI collective communication needs and existing IETF transport standards\",\n \"m\": 2,\n \"mn\": \"Early analysis phase - identifies problems and gaps but lacks detailed technical solutions or protocol specifications\",\n \"o\": 2,\n \"on\": \"Limited overlap with existing IETF work, addresses specific AI networking requirements not extensively covered elsewhere\",\n \"mo\": 2,\n \"mon\": \"Single revision draft with no clear working group engagement or community feedback visible\",\n \"r\": 4,\n \"rn\": \"Directly addresses critical AI infrastructure needs for distributed training and inference communication optimization\",\n \"c\": [\"ML traffic mgmt\", \"A2A protocols\"]\n}", "in_tok": 650, "out_tok": 254} {"draft_name": "draft-yang-nmrg-mcp-nm", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft explores applying the Model Context Protocol (MCP) to network management, proposing it as a framework to integrate AI capabilities into network operations and provide more agile architecture. It covers MCP server discovery, network element communication, and network exposure capabilities in IP networks using IETF technologies.\",\n \"n\": 3,\n \"nn\": \"Useful application of MCP to network management domain, though applying existing protocols to new domains is fairly incremental\",\n \"m\": 2,\n \"mn\": \"Early sketch stage - discusses applicability and use cases but lacks detailed protocol specifications or implementation details\",\n \"o\": 2,\n \"on\": \"Some conceptual overlap with other network management and AI integration drafts but focuses specifically on MCP application\",\n \"mo\": 2,\n \"mon\": \"Single revision with future date suggests early stage work without demonstrated community engagement yet\",\n \"r\": 4,\n \"rn\": \"Directly relevant to AI agent integration in networking, specifically addressing how MCP can enable AI-driven network management\",\n \"c\": [\"Autonomous netops\", \"A2A protocols\", \"Agent discovery/reg\", \"Data formats/interop\"]\n}", "in_tok": 563, "out_tok": 275} {"draft_name": "draft-mishra-oauth-agent-grants", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"DAAP defines a comprehensive OAuth 2.0-based authorization framework for AI agents acting on behalf of humans, incorporating cryptographic identity, consent flows, audit trails, and policy engines. The protocol addresses the critical gap of verifying human authorization for specific agent actions with real-time revocation and tamper-evident logging.\",\n \"n\": 4,\n \"nn\": \"Notable originality in combining OAuth patterns with agent-specific requirements like cascade revocation, hash-chained audit trails, and multi-agent delegation models\",\n \"m\": 4,\n \"mn\": \"Detailed 40-page specification with comprehensive protocol flows, token formats, and integration points, though lacks implementation examples\",\n \"o\": 2,\n \"on\": \"While building on OAuth 2.0 foundations, the agent-specific extensions and cryptographic identity integration represent a distinct approach\",\n \"mo\": 2,\n \"mon\": \"Single revision draft with 2026 date suggests early stage development without demonstrated working group or community engagement\",\n \"r\": 5,\n \"rn\": \"Core AI agent topic directly addressing fundamental authorization challenges for autonomous agent systems acting on human behalf\",\n \"c\": [\"Agent identity/auth\", \"Policy/governance\", \"Human-agent interaction\"]\n}", "in_tok": 729, "out_tok": 292} {"draft_name": "draft-mpsb-agntcy-slim", "model": "claude-sonnet-4-20250514", "raw": "{\"s\":\"SLIM is a protocol for real-time interactive AI applications that combines gRPC over HTTP/2/3 with secure messaging and group communication. It provides transport layer capabilities for agent protocols like A2A and MCP with connection management, stream multiplexing, and MLS encryption.\",\"n\":2,\"nn\":\"Incremental combination of existing technologies (gRPC, HTTP/2/3, MLS) for AI agent transport\",\"m\":3,\"mn\":\"Defines protocol mechanisms and integration approach but lacks detailed specifications and implementation examples\",\"o\":4,\"on\":\"Significant overlap with existing gRPC, HTTP transport, and MLS specifications - primarily integrates existing protocols\",\"mo\":2,\"mon\":\"Single revision from February 2026 with no clear working group adoption or community activity\",\"r\":4,\"rn\":\"Directly relevant to AI agent infrastructure, specifically addressing transport needs for agent-to-agent communication\",\"c\":[\"A2A protocols\",\"Data formats/interop\"]}", "in_tok": 591, "out_tok": 215} {"draft_name": "draft-wahl-scim-agent-schema", "model": "claude-sonnet-4-20250514", "raw": "{\n \"s\": \"This draft extends SCIM identity management to support AI agents by defining a JSON schema for representing agentic identities. It enables AI agents to be provisioned, authenticated, and authorized across cloud services using existing SCIM infrastructure.\",\n \"n\": 3,\n \"nn\": \"Useful contribution applying established identity management patterns to the emerging need for AI agent identity representation\",\n \"m\": 2,\n \"mn\": \"Early sketch level - defines basic schema concepts but lacks detailed protocol mechanisms and implementation guidance\",\n \"o\": 2,\n \"on\": \"Minor similarities to general identity management work but addresses the specific domain of AI agent identities\",\n \"mo\": 2,\n \"mon\": \"Single revision from 2026 with no indication of working group adoption or community feedback\",\n \"r\": 4,\n \"rn\": \"Directly addresses a core infrastructure need for AI agent deployment - identity management and cross-domain authentication\",\n \"c\": [\"Agent identity/auth\", \"Data formats/interop\"]\n}", "in_tok": 553, "out_tok": 240}