OpenAI, Anthropic and Block have placed their flagship agent-building tools under a new neutral body, the Agentic AI Foundation (AAIF), housed at the Linux Foundation and bank-rolled by more than a dozen cloud and finance giants.
Announced on 9 December, the initiative intends to create open, shared standards for connecting, instructing and running autonomous AI agents before the market fragments into proprietary walled gardens.
Anthropic is contributing its Model Context Protocol (MCP), already used by 10,000 public servers and baked into ChatGPT, VS Code and Microsoft Copilot. OpenAI is handing over AGENTS.md, a markdown convention adopted by 60,000 GitHub projects that lets agents read repository-specific instructions. Block is donating Goose, an open-source framework that links large language models to plug-in tools.
“Bringing these projects together ensures they grow with the transparency and stability that only open governance provides,” said Linux Foundation executive director Jim Zemlin. Technical road-maps will be set by elected steering committees; no single member can veto changes.
Manik Surtani, Block’s head of open source, framed the move as a choice “between closed systems that benefit a few or open standards that benefit everyone.”
The AAIF will coordinate interoperability, safety patterns and best-practice documentation, and will be funded by membership dues. Platinum backers include Amazon Web Services, Google, Microsoft, Bloomberg and Cloudflare.
Founders say the immediate payoff for developers should be fewer custom connectors and more predictable agent behaviour inside security-conscious enterprises. The long-term goal is an open agent ecosystem akin to the web standards that underpinned the modern internet.