The mission of the AI Agent Protocol Community Group is to develop open, interoperable protocols that enable AI agents to discover, identify, and collaborate efficiently across the Web. As AI agents increasingly participate in Web-based activities, there is a growing need for standardized mechanisms to support secure, reliable, and scalable interactions. This group aims to establish the technical foundations for the emerging Agentic Web. The AI Agent Protocol Community Group would like to bring together industry and academia stakeholders interested in standardizing agent communication.
Scope of Work
This Community Group will explore and define:
Inter-agent communication protocols: Mechanisms that allow agents to discover one another, exchange intent and capability information, negotiate roles, and dynamically establish or dissolve collaborations in an open Web environment.
Agent identity models: An identity framework for AI agents, based on open standards, to support secure, interoperable authentication between agents across domains.
Standardized metadata formats: Structured descriptions of agent capabilities, interfaces, goals, and states, based on Web-native formats to enable automated reasoning, composition, and orchestration of agent behavior.
Security and privacy mechanisms: Cross-origin communication security models for agents, including authentication, authorization, verifiable credential-based trust, and end-to-end encryption, to safeguard the integrity and confidentiality of agent interactions.
Protocol interoperability: Compatibility layers and best practices that allow agent-based systems to leverage existing Web protocols and standards for integration and collaboration.
This group will publish Specifications.
This group welcomes participation from AI platform developers, protocol designers, browser vendors, Web application developers, academic researchers, and other stakeholders interested in advancing the open standards that will enable a trusted, collaborative Web of agents.
Who Should Join
We welcome participation from individuals and organizations in the following areas:
Developers of AI platforms and agent-based applications
Enterprise software and infrastructure providers
Academic and research institutions
Standards bodies and open source organizations
Policymakers and regulatory agencies
Join the AI Agent Protocol Web Community Group to co-develop open technologies that enhance the interoperability and security of AI agents—and help build a trustworthy, collaborative, and global agent network for the Web.
Note: Community Groups are proposed and run by the community. Although W3C hosts these
conversations, the groups do not necessarily represent the views of the W3C Membership or staff.
The AI Agent Protocol Community Group now holds regular biweekly online meetings, and we warmly welcome your participation. You can find the meeting schedule and joining details on our W3C calendar: 👉 https://www.w3.org/groups/cg/agentprotocol/calendar/
Meeting summaries are available on GitHub meetings for those who wish to catch up asynchronously. If the current meeting time is inconvenient for your time zone, please let us know—we’re happy to adjust one of the alternate meeting slots to accommodate global participation better.
Our ongoing work is hosted on GitHub, focusing on two core documents:
The mission of the AI Agent Protocol Community Group is to develop open, interoperable protocols that enable AI agents to discover, identify, and collaborate efficiently across the Web. As AI agents increasingly participate in Web-based activities, there is a growing need for standardized mechanisms to support secure, reliable, and scalable interactions. This group aims to establish the technical foundations for the emerging Agentic Web. The AI Agent Protocol Community Group would like to bring together industry and academia stakeholders interested in standardizing agent communication.
Scope of Work
This Community Group will explore and define:
Inter-agent communication protocols: Mechanisms that allow agents to discover one another, exchange intent and capability information, negotiate roles, and dynamically establish or dissolve collaborations in an open Web environment.
Agent identity models: An identity framework for AI agents, based on open standards, to support secure, interoperable authentication between agents across domains.
Standardized metadata formats: Structured descriptions of agent capabilities, interfaces, goals, and states, based on Web-native formats to enable automated reasoning, composition, and orchestration of agent behavior.
Security and privacy mechanisms: Cross-origin communication security models for agents, including authentication, authorization, verifiable credential-based trust, and end-to-end encryption, to safeguard the integrity and confidentiality of agent interactions.
Protocol interoperability: Compatibility layers and best practices that allow agent-based systems to leverage existing Web protocols and standards for integration and collaboration.
This group will publish Specifications.
This group welcomes participation from AI platform developers, protocol designers, browser vendors, Web application developers, academic researchers, and other stakeholders interested in advancing the open standards that will enable a trusted, collaborative Web of agents.
Who Should Join
We welcome participation from individuals and organizations in the following areas:
Developers of AI platforms and agent-based applications
Enterprise software and infrastructure providers
Academic and research institutions
Standards bodies and open source organizations
Policymakers and regulatory agencies
Join the AI Agent Protocol Web Community Group to co-develop open technologies that enhance the interoperability and security of AI agents—and help build a trustworthy, collaborative, and global agent network for the Web.
This is a community initiative. This group was originally proposed on 2025-05-08 by
gaowei chang. The following people supported its creation: Ruoxi Ran, Song XU, gaowei chang, Gavin Sze and Jomy Wu.
W3C’s hosting of this group does not imply endorsement of the activities.
w3c-cg/ai-agent-protocol
