The race to standardize AI agents just took a sharp turn. Anthropic, OpenAI, and Block have teamed up under the Linux Foundation to launch the Agentic AI Foundation (AAIF) — a new industry-backed group designed to create open standards for the fast-growing world of agentic AI.
It’s a rare moment of cooperation in a field defined by rivalry. And it arrives at a time when agents are quickly shifting from fun demos to core infrastructure.
A New Power Bloc for Agent Standards
Agentic AI — systems that act, plan, and use tools on their own — has exploded. But the ecosystem is messy. Every company built its own protocols. Every developer rolled out their own tool layer. None of it talked to anything else.
AAIF wants to end the chaos.
The foundation is structured as a neutral, open governance hub. Think the Kubernetes moment for AI agents: one rulebook, many contributors, and infrastructure that anyone can build on.
And the contributors aren’t small names.
Today, the @linuxfoundation launched the Agentic AI Foundation with Anthropic and OpenAI donating MCP and https://t.co/p7zkNkh0LK as founding projects.
— Docker (@Docker) December 9, 2025
Docker is joining alongside Anthropic, Google, OpenAI, Cloudflare, and others.
Open standards won the application layer.… pic.twitter.com/vtahMkoNsr
Anthropic Donates MCP, Now Everywhere
Anthropic is handing over the Model Context Protocol (MCP) — a framework already running across 10,000+ servers and powering integrations inside ChatGPT, Microsoft Copilot, and dozens of emerging developer tools.
MCP became a de-facto standard in under a year. Moving it under the Linux Foundation ensures it won’t become a single-company asset. It becomes a shared backbone instead.
GitHub, which helped shape the protocol early on, called the move “a massive moment” for the agent ecosystem.
OpenAI Adds AGENTS.md — A Spec Used in 20,000 Projects
OpenAI is also donating AGENTS.md, a minimalist but influential spec that defines how agents should manage tasks, tools, retries, and context.
Developers embraced it fast. More than 20,000 projects already rely on it.
OpenAI says the goal is open interoperability — not a new walled garden. That’s a notable shift for a company often criticized for siloed products.
Block Contributes Goose for Secure Workflows
Block is bringing Goose, its secure agent framework built for finance, compliance, and regulated workflows.
This matters because financial agents are one of the hottest — and riskiest — emerging use cases. Goose adds a hardened, audit-friendly architecture the industry has been missing.
Together, MCP, AGENTS.md, and Goose form the first wave of AAIF’s technical stack.
A Large Coalition Steps In
The founding trio isn’t alone. Google, Microsoft, AWS, Docker, Cloudflare, and GitHub have all joined the coalition, offering engineering support and advisory roles.
Docker summed up the mood: “Open standards won the application layer.” Now, they want the same for agents.
If this momentum holds, AAIF could become an industry default before competitors even agree on alternatives.
Why This Matters Now
The timing isn’t accidental. Agents are moving into browsers, operating systems, productivity suites, and developer workflows. Companies want reliability. Enterprises want stability. Regulators want oversight.
Fragmentation makes all three harder.
AAIF’s pitch hits the core concerns:
- A single interface for tools and data
- Reduced vendor lock-in
- Shared security assumptions
- Faster development cycles
It’s the infrastructure piece the agent ecosystem has been missing.
The Bigger Picture: A Rare Industry Truce
What makes this launch striking is the cooperation between rivals. OpenAI and Anthropic usually compete at the model layer. Block and Microsoft rarely align on architecture. Cloudflare and AWS compete everywhere.
And yet, here they are — building a shared foundation.
The message is clear: agent interoperability is too important to leave fragmented.
If the ecosystem consolidates around open standards, developers and enterprises benefit first. And the companies behind those standards gain influence without owning the stack outright.
What Comes Next
AAIF will begin forming working groups around tooling, reference implementations, and safety frameworks. Expect rapid movement. The agent wave is expanding faster than any previous developer trend, and companies want common rails before adoption outpaces regulation.
If the foundation succeeds, agentic AI could leap from a collection of prototypes into a unified, cross-platform ecosystem — one that feels as cohesive as the early container era.
Conclusion
The AI giants just agreed on something big: a shared future for agentic AI. With three major frameworks donated and cloud giants signing on, AAIF could become the backbone of next-generation automation.
The open-standards era for AI agents has officially begun — and the stakes couldn’t be higher.