The IETF Agent Communication Protocols working group develops a common foundation for maintaining dialogs among users, AI agents, and tools across platforms and vendors.
Our work focuses on dialog identifiers, lifecycle management, and propagation of dialog context across intermediaries, trust boundaries, and modalities, with bindings to existing IETF protocols. Here, dialog context means the identifiers and state needed to correlate interactions and maintain continuity; it excludes prompts, conversation memory, and other model content.
The current charter defines three deliverables:
- Agentic Dialog Management Protocol: standards-track specifications for dialog correlation, continuity, lifecycle management, and protocol bindings, addressing security and privacy risks.
- Reference Architecture: an informational description of how existing identity, authentication, authorization, and encryption building blocks support the protocol.
- Use Cases and Requirements: an informational description of the scenarios, requirements, and deployment models that guide the work.
The scope complements existing application protocols such as MCP and A2A. New transport protocols, agent reasoning and planning, model behavior, user interfaces, and audit data structures are outside the charter's scope.
- Mailing list: agentproto@ietf.org · Subscribe · Archives
- Slack: Join the IETF workspace for AgentProto discussion.
- Documents: Internet-Drafts and document status
- Working group information: Charter, milestones, and chairs
Use GitHub issues and pull requests to contribute to documents. Bring substantive discussions and proposed decisions from GitHub and Slack to the mailing list so the whole working group can participate and establish consensus.
Contributions here, including issues, comments, commits, and pull requests, are part of the IETF Standards Process. Participation is subject to the IETF Note Well, including BCP 78, BCP 79, and the IETF Trust Legal Provisions.
