A2A joins MCP in the Agentic AI Foundation
The A2A protocol has moved to the Agentic AI Foundation, where MCP already lives. What one home for both agent protocols means, and what it does not solve.
The two protocols that matter most for AI agents in integration now have the same home. On 17 August, it became known that the Agent2Agent protocol, A2A, is moving to the Agentic AI Foundation (AAIF), the Linux Foundation body that already hosts the Model Context Protocol.
What happened
A2A was created by Google and handed to the Linux Foundation in June 2025. It has so far been run as a separate Linux Foundation project. Now it becomes a hosted project of AAIF.
AAIF was formed in December 2025. According to Axios, it has grown from fewer than 40 members to more than 250, among them Google, Microsoft, Amazon, Anthropic, OpenAI, Bloomberg, Shopify and Block. Besides MCP and now A2A, the foundation hosts Block’s agent runtime goose, OpenAI’s AGENTS.md convention and the AgentGateway proxy, according to Forbes.
The two protocols do different jobs. MCP connects an agent to tools and data. A2A lets independent agents find each other, hand over tasks and work together, across vendors and frameworks.
Mazin Gilbert, executive director of AAIF, told Axios that companies do not want just one protocol, they want the whole stack to be open.
Why it matters
For integration teams, the most important effect is predictability. Both protocols are now under neutral governance in the same foundation, with the same kind of processes. That makes it easier to build on them for the long term, and harder for a single vendor to pull them in its own direction.
It also removes some of the fear of a standards war. When MCP and A2A were run in different places, many asked whether they would compete. Now they are placed side by side as parts of the same open stack.
What it does not solve
Forbes points out that this is coordination, not a change of control. Each protocol keeps its own maintainers, specification process and release schedule. And the move does not standardise how companies should authorise agents. Who may call which agent, with which rights, and how it is logged, is still up to each company and its platforms.
That is where integration platforms and gateways come in. The protocols define how agents talk. Policies, identity and audit are still the job of the platform the traffic passes through.
What to ask
- Does your platform support both? Check whether your iPaaS and API gateway handle both MCP and A2A, and which versions.
- Where are the agent policies? Decide where the rules for agent-to-agent calls are set, since the protocol does not do it for you.
- Who follows the standards work? Give someone responsibility for following changes to both protocols, now that they move in the same forum.
Sources
This post was written with AI assistance and reviewed by the editor before publishing.