Linux Foundation Launches Agentic AI Foundation to Govern MCP and Open Agent Standards, With Anthropic, OpenAI, and Block as Co-Founders
The Linux Foundation established AAIF on December 9, 2025, placing MCP, goose, and AGENTS.md under neutral open-source governance with eight platinum members including AWS, Google, and Microsoft.
Overview
The Linux Foundation on December 9, 2025, announced the formation of the Agentic AI Foundation (AAIF), a new directed fund intended to provide neutral open-source governance for the tools and protocols that underpin autonomous AI agents. Three of the AI industry’s most prominent companies — Anthropic, Block, and OpenAI — co-founded the organization, each donating a flagship project to serve as the foundation’s initial technical anchors.
The move marks a significant shift in how the major AI laboratories are approaching the infrastructure layer of agentic AI: rather than competing on protocol standards, they are pooling governance of core interoperability specifications under a neutral steward with a track record in critical open-source infrastructure.
The Three Founding Projects
Model Context Protocol (MCP)
Anthropics donated the Model Context Protocol — a universal standard it originally open-sourced in November 2024 — to serve as the foundation’s flagship project. MCP defines how AI models connect to external tools, data sources, and applications. As reported by TechCrunch, the protocol had already achieved broad adoption before the AAIF announcement, with over 10,000 published MCP servers and integration into Claude, Cursor, Microsoft Copilot, Gemini, Visual Studio Code, and ChatGPT.
According to the Linux Foundation press release, the November 2025 MCP specification release introduced asynchronous operations, statelessness, and official SDKs across major programming languages, reaching 97 million monthly SDK downloads by the time of the AAIF’s formation.
goose
Block, the financial technology company founded by Jack Dorsey, contributed goose — an open-source, local-first AI agent framework that combines language models with extensible MCP-based tool integration. Released in early 2025, goose emphasizes structured, reliable workflows for building agentic applications and is designed to run locally, reducing dependency on cloud infrastructure for agent execution.
AGENTS.md
OpenAI donated AGENTS.md, a markdown-based convention that gives AI coding agents consistent, project-specific guidance across different repositories and toolchains. By placing a standardized file at the root of a code repository, developers can tell any compliant AI agent how to work within that project — what tools to use, what conventions to follow, and what pitfalls to avoid. The standard had been adopted by more than 60,000 open-source projects and agent frameworks by the time of the AAIF’s launch, including GitHub Copilot, Cursor, Codex, Devin, Gemini CLI, and VS Code.
Governance and Membership
The AAIF is structured as a directed fund under the Linux Foundation, the nonprofit that governs Kubernetes, the Linux kernel, PyTorch, and dozens of other critical open-source projects. According to the foundation’s announcement, eight organizations joined at the platinum membership tier: Amazon Web Services, Anthropic, Block, Bloomberg, Cloudflare, Google, Microsoft, and OpenAI. Gold and silver tiers include more than 40 additional members, among them Salesforce, Snowflake, Docker, IBM, Oracle, Hugging Face, Pydantic, and Zapier.
Jim Zemlin, executive director of the Linux Foundation, stated in the announcement that “AI is entering a new phase as conversational systems shift to autonomous agents that can work together,” and that the foundation would ensure critical agentic infrastructure “evolves transparently, collaboratively, and in ways that advance the adoption of leading open source AI projects.”
Day-to-day governance of each project — including MCP — remains with existing maintainers. Obot.ai also donated the MCP Dev Summit events and associated podcast to the AAIF, transferring community infrastructure alongside the technical projects.
What the AAIF Is Intended to Solve
The agentic AI ecosystem has developed rapidly, but without a common layer for interoperability, individual vendors have released proprietary tool-calling formats, agent communication protocols, and context-passing conventions. The fragmentation risk mirrors patterns seen in earlier technology cycles — such as container orchestration before Kubernetes consolidated governance, or JavaScript tooling before standards bodies intervened.
As TechCrunch noted at the time of the announcement, the goal is to prevent the agent ecosystem from becoming locked into competing, incompatible proprietary systems. By placing MCP and AGENTS.md under neutral governance with membership from every major AI platform vendor, the AAIF is positioned to arbitrate disputes over protocol evolution without any single company exercising unilateral control.
MCP Dev Summit North America 2026
The AAIF announced on February 24, 2026, that the MCP Dev Summit North America will be held in New York City on April 2–3, 2026, with more than 95 conference sessions drawn from MCP co-founders, maintainers, contributors, and production deployers. The program covers protocol evolution, conformance testing, security research, production deployment lessons, and scalable agent system design.
Notable sessions include “One Spec, Ten SDKs, Zero Excuses: Conformance Testing MCP” by Paul Carleton of Anthropic; “When MCP Isn’t Enough: Product Decisions Behind Scalable Agent Systems” by Cansu Berkem of Datadog; “Mix-Up Attacks in MCP: Multi-Issuer Confusion and Mitigations” by Emily Lauber of Microsoft; and “MCP at 18 Months: Protocols, Patterns, and What We Didn’t See Coming” by Shaun Smith of Hugging Face.
What Remains Unclear
Several questions about the AAIF’s long-term role remain open. The foundation has not yet published a formal process for adjudicating protocol disputes among founding members, which could become significant if Anthropic’s preferred MCP evolution conflicts with OpenAI’s or Microsoft’s priorities. The AAIF also has not yet specified a timeline for MCP 1.0 stabilization, leaving enterprise adopters uncertain about breaking-change risk.
It is also unclear how AGENTS.md and MCP will formally interoperate at the specification level: while both are under the same governance umbrella, they were designed independently, and whether a unified agent context standard will emerge from the foundation is an open question.
Analysis
The AAIF’s formation follows a pattern familiar in open-source governance: a dominant incumbent protocol — in this case MCP, which achieved critical mass in roughly 12 months — is transferred to a neutral body to forestall a challenger ecosystem forming around a competing standard. By inviting OpenAI into co-governance before a rival agent protocol gained traction, Anthropic effectively locked in MCP as the default while accepting a governance constraint in exchange.
For developers, the practical near-term implication is that MCP investments are lower risk than they were when the protocol was controlled by a single company. The AAIF structure also creates a formal venue for the conformance testing and security research that production-scale MCP deployments will require — a gap the April summit’s session lineup is explicitly designed to address.