Google's A2A protocol joins Anthropic's MCP under one neutral foundation — AI agent standards just consolidated
On 17 August 2026, Google's Agent2Agent (A2A) protocol became a hosted project of the Linux Foundation's Agentic AI Foundation alongside Anthropic's Model Context Protocol (MCP), which has grown from under 40 members at launch to more than 250 in eight months — a sign the tools for building multi-agent software are settling on shared, vendor-neutral standards.
29 August 2026
On 17 August 2026, Google’s Agent2Agent (A2A) protocol — its standard for how one AI agent asks another to complete a task and hand back a result — became a hosted project of the Agentic AI Foundation (AAIF) at the Linux Foundation. It joins Anthropic’s Model Context Protocol (MCP), which standardises how an agent connects to a database, an API or a file system, and which has been part of AAIF since the foundation launched last December. Membership has grown from under 40 organisations at launch to more than 250 in eight months, spanning AWS, Google, Microsoft, OpenAI and Anthropic.
The two protocols solve different problems and each keeps its own maintainers and release schedule — this isn’t a merger of the specifications, it’s a merger of governance. But that distinction matters less than the headline: the two protocols a founder is most likely to hit when scoping any AI-agent feature — how an agent talks to your systems, and how it talks to other agents — now sit under the same neutral, vendor-independent umbrella, backed by every major AI lab and cloud provider rather than owned by one of them.
Why vendor-neutral governance changes the buy decision
Standards that live inside one vendor’s product roadmap carry switching risk — build against them and you’re betting the vendor keeps supporting your use case. Standards governed by a neutral foundation with 250+ members carry a different kind of confidence: no single company can deprecate them out from under you, and competing implementations tend to converge on compatibility rather than fragment. For any team building custom software with AI-agent features — automated workflows, tool-using assistants, systems that call out to other systems — this is the point where “which protocol should we build on” stops being a risky bet and starts being closer to a settled question, the way HTTP or SQL are settled questions.
So what
If your project involves agents that need to call your APIs, your data, or each other, it’s worth knowing whether the approach you’re building on rests on a standard this broad or on a single vendor’s proprietary format — the difference shows up later as either a smooth integration or a costly rebuild. We factor protocol choice into scoping AI-agent features as a matter of course. See our AI-assisted development work, or get in touch if you’re planning a build with agentic components.