Agent Communication Protocols
bofreq-krishnan-agent-communication-protocols-02
| Document | Type | Approved BOF request | |
|---|---|---|---|
| Title | Agent Communication Protocols | ||
| Last updated | 2026-06-11 | ||
| State | Approved | ||
| Editor | Suresh Krishnan | ||
| Responsible leadership | |||
| Send notices to | (None) |
Name: Agent Communication Protocols (agentproto)
Description
An AI agent is an autonomous, adaptive intelligent software system that uses AI models to complete a specific task. AI Agents often interact with users via chat or voice, performing tasks based on the flow of the conversation. To complete tasks on behalf of a human user or another AI agent, they can independently make decisions, execute actions, and interact with other AI agents and tools.
With the expansion of communication over the Internet between AI Agents and external resources that can be tools or other AI Agents, reliable communications across platforms and vendors becomes increasingly important. The role of the this effort is to facilitate such interoperability so that tools and agents can be provided by multiple vendors. The Agent Communication Protocols effort will work on defining frameworks and protocol building blocks for enabling interoperability for AI Agent applications across the Internet.
Required Details
- Status: WG forming
- Responsible AD: Charles Eckel
- BOF proponents:
Cullen Jennings <fluffy@iii.ca>
Jonathan Rosenberg <jdrosen@jdrosen.net>
Suresh Krishnan <suresh.krishnan@gmail.com>
Dapeng Liu <max.ldp@alibaba-inc.com>
Zaheduzzaman Sarker <zaheduzzaman.sarker@nokia.com>
Lionel Morand <lionel.morand@huawei.com>
Kehan Yao <yaokehan@chinamobile.com>
Arashmid Akhavain <arashmid.akhavain@huawei.com>
Hesham Moussa <hesham.moussa@huawei.com>
- Number of people expected to attend: 200
- Length of session (1 or usually 2 hours): 2 hours
- Conflicts (whole Areas and/or WGs)
- Chair Conflicts: aipref, iabopen, oauth
- Technology Overlap: dawn(BoF), oauth, moq, quic, webtrans, wimse
- Key Participant Conflict: TBD
Information for IAB/IESG
To allow evaluation of your proposal, please include the following items:
- Any protocols or practices that already exist in this space:
Yes. There are several protocols and frameworks that exist in this space. Some of the key protocols and frameworks include the Model Context Protocol (MCP), the Agent2Agent (A2A) protocol, and the AGNTCY framework.
Agentic AI Foundation (Home of MCP)
https://aaif.io
Agent2Agent (A2A) Protocol
https://a2a-protocol.org/latest/
https://github.com/a2aproject/A2A
AGNTCY
https://agntcy.org
https://github.com/agntcy
Model Context Protocol (MCP)
https://modelcontextprotocol.io/
https://github.com/modelcontextprotocol
- Which (if any) modifications to existing protocols or practices are required:
Yes. Here are some of the modifications that are anticipated:
-
This effort is expected to work on a protocol for creation and maintenance of long-lived sessions between AI agents, or between AI agents and tools. These sessions allow for the bidirectional exchange of data, including model context, tool call results, and chat messages. It is anticipated that this protocol will utilize modern IETF application transfer protocols, such as QUIC, WebTransport, WebRTC and MOQ, and might require some modifications or additions.
-
This effort is expected to work on an agent-specific generic communication framework that will enable an AI Agent to select and collaborate with other AI agents/tools, allow multi-modal collaboration, create an independent identity, obtain and exchange access tokens with fine-grained, behavior-driven scopes bound to the specific operations that it is permitted to perform on behalf of the user.
The framework includes work about documenting how existing and emerging mechanisms are composed and applied in anticipated AI agent scenarios/use cases. Any extensions to existing IETF protocol mechanisms required to support the agent communication framework are expected to be developed within the existing working groups where the respective protocols are developed.
- Which (if any) entirely new protocols or practices are required:
This effort is expected to work on a protocol for creation and maintenance of long-lived sessions between AI agents, or between AI agents and tools. These sessions allow for the bidirectional exchange of data, including model context, tool call results, and chat messages. It is anticipated that this protocol will utilize modern IETF application transfer protocols, such as QUIC, WebTransport, WebRTC and MOQ, based on the anticipated use cases.
- Open source projects (if any) implementing this work:
https://alibaba.github.io/open-agent-auth/
Open source work including existing protocols that are related to this work
Agent2Agent (A2A) Protocol
https://github.com/a2aproject/A2A
AGNTCY
https://github.com/agntcy
Model Context Protocol (MCP)
https://github.com/modelcontextprotocol
- Co-ordination with Open source work
The proponents recognize the importance of close coordination between IETF standardization work and the open source projects listed, several of which are hosted by or affiliated with the Linux Foundation. Several of the proponents have been working closely with these open source projects in order to gather requirements and preferences, in order to improve the ability of the IETF to define the necessary building blocks that these open source projects can adopt and deploy. We expect to have ongoing information exchange and co-ordination with these projects by way of common participants between the projects and the IETF working group(s) in this space.
Agenda
- Items, Internet-Drafts, speakers, timing
- Or a URL
Links to the mailing list, draft charter if any (for WG-forming BoF), relevant Internet-Drafts, etc.
- Mailing List: https://www.ietf.org/mailman/listinfo/agent2agent
- Draft charter: https://github.com/jdrosen/aiproto-wg/blob/main/Charter_Merge.md
- Relevant Internet-Drafts:
https://datatracker.ietf.org/doc/draft-rosenberg-aiproto-framework/
https://datatracker.ietf.org/doc/draft-jennings-ai-mcp-over-moq/
https://datatracker.ietf.org/doc/draft-klrc-aiagent-auth/
https://datatracker.ietf.org/doc/draft-liu-agent-operation-authorization/
https://datatracker.ietf.org/doc/draft-liu-agent-protocol-over-moq/
https://github.com/maxpassion/IETF-Agent-Operation-Authorization-draft
https://datatracker.ietf.org/doc/draft-agentic-ai-usecases-requirements/
https://datatracker.ietf.org/doc/draft-yao-catalist-problem-space-analysis/