The Agent2Agent protocol is moving into the Agentic AI Foundation, the Linux Foundation body that already hosts the Model Context Protocol, Block’s goose runtime, OpenAI’s AGENTS.md convention and the agentgateway proxy. AAIF announced the move on August 17.
The framing of this news needs care, because Google handed A2A to neutral governance more than a year ago, donating the specification, SDKs and tooling to a Linux Foundation project in June 2025 alongside AWS, Cisco, Microsoft, Salesforce, SAP and ServiceNow. What changed this month is which foundation hosts the work, not who controls it.
For architects evaluating multi-agent systems, this is a coordination story rather than a merger. Two of the most prominent interoperability protocols in the open agent stack now sit under one foundation, while each keeps its own maintainers, specification process and release schedule.
How A2A Actually Works
At its core, A2A is a discovery and delegation mechanism. An agent publishes an agent card, a structured description of what it can do and how to reach it. Other agents read that card and hand off work without a human brokering the connection. MCP standardizes how an agent reaches a database, an API or a file system. A2A standardizes how one agent asks another to complete a task and return the result.
Google launched A2A in April 2025, and the protocol reached v1.0 in March 2026. AAIF says that release added multi-protocol bindings and version negotiation, multi-tenancy support and cryptographically signed agent cards. A signed card authenticates the card’s identity and metadata against a trusted signing key. That guarantee holds only as far as the recipient can establish that the key belongs to the organization being asserted.
How A2A Reached AAIF
IBM merged its Agent Communication Protocol into A2A in August 2025, eliminating one significant competing specification, though Cisco’s AGNTCY still addresses overlapping ground in discovery and identity. AAIF itself formed only in December, so A2A is joining a young organization.
Adoption is broader than the usual protocol announcement, though the figures come from the foundation and its vendors rather than independent measurement. The Linux Foundation reported more than 150 supporting organizations and production deployments across several industries. AAIF cites Huawei’s use of A2A between Celia, its operating system assistant, and in-app agents on HarmonyOS. It also lists support in Google Cloud, Azure AI Foundry and AWS Bedrock AgentCore.
What The Transfer Does Not Solve
A shared foundation settles who stewards the specifications. It does not standardize enterprise authorization policy. A2A already carries normative requirements here, since a server must perform authorization checks and scope results to the caller’s authorized boundaries. The policy model itself is deliberately left open, so what a partner’s agent may touch once a task crosses an organizational boundary still lands on the enterprise.
Implementation depth varies widely across the platforms claiming support, since hosting an A2A endpoint is not the same as supporting protocol bindings, version negotiation, multi-tenancy and signed cards. Consensus governance can also add coordination overhead, although whether that slows A2A’s release cadence under AAIF remains to be seen.
What Enterprise Buyers Should Ask
The first question for buyers is conformance. Which A2A revision does the vendor implement, which bindings does it support, and does it negotiate versions with older clients? The second question covers trust and isolation. Does the platform reject unsigned agent cards, how does it establish that a signing key belongs to the asserted organization, and what isolates one tenant’s agents from another’s?
The third question, and the one buyers skip most often, is accountability. When an enterprise agent delegates work to a partner’s agent, which system enforces the access boundary, and does the exchange leave an audit record a compliance team can read?
With both protocols now governed in the same way and in the same place, the MCP and A2A communities gain a venue to coordinate in. Enterprises also carry less risk that either specification turns on one vendor’s roadmap. The specifications remain separate, so the work of making agents interoperate still belongs to the teams deploying them.
