Anthropic hands MCP over to the Linux Foundation

When sworn rivals suddenly decide to hold hands, it usually means one of two things: an extinction-level threat, or a massive infrastructure pie nobody can secure alone.

Anthropic and OpenAI just teamed up with Block to co-found the Agentic AI Foundation (AAIF) under the neutral umbrella of the Linux Foundation. Backing them up is an absolute tech heavyweight lineup: Google, Microsoft, AWS, Cloudflare, and Bloomberg.

And nobody showed up empty-handed. Anthropic donated its Model Context Protocol (MCP), OpenAI handed over AGENTS.md, and Block contributed its open-source agent framework, goose.

Giving up control to set the standard

MCP has basically become the USB port of the agentic era. Introduced just a year ago by Anthropic, it connects AI models to external tools and corporate data.

The numbers behind it are wild: over 10,000 active public MCP servers are running today, powering everything from scrappy dev projects to Fortune 500 setups. Its Python and TypeScript SDKs clear 97 million monthly downloads, supported natively by ChatGPT, Cursor, Gemini, and VS Code.

OpenAI’s donation, AGENTS.md, solves the other half of the puzzle. It is essentially a project guide sitting right next to README.md, giving coding agents clear ground rules on testing, builds, and code styles.

Since its release in August 2025, more than 60,000 repositories and agent engines—from Devin to GitHub Copilot—have adopted it. OpenAI noted that its Codex tooling alone has helped merge over 2 million public pull requests on GitHub.

Why it matters

We think this is one of the smartest and most calculated strategic moves the AI industry has seen all year.

For developers and enterprises, fragmentation was turning into an absolute nightmare. Nobody wanted to write custom glue code every time they switched between Claude and GPT. With MCP and AGENTS.md housed in a vendor-neutral home, building an agent tool once means it will work everywhere, making it an absolute no-brainer to adopt right now.

For the frontier labs, giving up unilateral control is a masterclass in platform defense. By handing these protocols to the Linux Foundation, Anthropic and OpenAI just locked in their designs as the global baseline. Any rival hoping to push a proprietary protocol is now effectively shut out before they even start.

The real test, of course, comes next. The Linux Foundation knows how to keep Kubernetes and Linux running peacefully, but watching Anthropic and OpenAI fight over future protocol roadmaps inside a committee is going to be the real show.