Block, Anthropic, and OpenAI Co-Found the Agentic AI Foundation — MCP and Goose Go Neutral
Block, Anthropic, and OpenAI have co-founded the Agentic AI Foundation (AAIF) under the Linux Foundation, donating three anchor projects: Anthropic's Model Context Protocol, Block's Goose open-source agent framework, and OpenAI's AGENTS.md specification. The foundation held its first MCP Dev Summit in New York in April 2026.
Original sourceThree of the biggest names in AI infrastructure announced the formation of the Agentic AI Foundation (AAIF) this week, a neutral governance body under the Linux Foundation dedicated to open standards and tooling for AI agent frameworks. The founding members — Block, Anthropic, and OpenAI — contributed three anchor projects: Anthropic's Model Context Protocol (MCP), Block's Goose open-source agent framework, and OpenAI's AGENTS.md specification for agent behavior.
The AAIF is structured as an industry-neutral foundation, analogous to what the Linux Foundation did for cloud-native computing through the CNCF. The goal is to prevent any single company from controlling the standards underlying agent-to-tool communication, agent behavior specifications, and agent hosting infrastructure — concerns that have grown louder as MCP has accelerated from Anthropic-specific protocol to something every major LLM client is implementing.
Goose, Block's open-source AI agent framework built in Rust, is the most mature of the donated projects. It already supports 15+ LLM providers, has an active plugin ecosystem, and shipped v1.28.0 this week with adversarial agent support and Claude adaptive thinking. Its move to the AAIF GitHub organization is immediately visible — it's no longer under Block's corporate GitHub umbrella.
MCP's inclusion is significant: the protocol has spread faster than almost any AI infrastructure standard in recent memory, with Claude Desktop, Cursor, Windsurf, VS Code, and Cline all supporting it natively. Moving governance to a neutral foundation reduces the risk that competitors refuse adoption because it's an "Anthropic protocol." AGENTS.md from OpenAI provides a complementary specification for agent capability declaration and behavior bounds.
The AAIF held its inaugural MCP Dev Summit in New York in April, drawing representatives from multiple AI companies. Industry reaction has been largely positive — developers who've been building on MCP are relieved to see governance formalized, though some skeptics note that "foundation" governance can slow iteration. The founding members have committed to maintaining open participation and avoiding proprietary forks of the core specifications.
Panel Takes
The Builder
Developer Perspective
“MCP moving to a neutral foundation is exactly what it needed to become infrastructure-grade. When the protocol sits under one company, enterprise procurement is harder and competitors have an excuse not to support it fully. AAIF governance removes both objections. Goose getting the same treatment makes the whole stack more credible for production agent deployments.”
The Skeptic
Reality Check
“Foundation governance sounds neutral but often just formalizes the power of founding members. Anthropic, Block, and OpenAI will set the agenda for MCP and AGENTS.md regardless of the org chart. The real test is whether Google, Meta, and Microsoft join as genuine participants — or whether AAIF becomes another foundation that sounds open but moves to the rhythm of three companies.”
The Futurist
Big Picture
“This is the CNCF moment for AI agents. When Kubernetes moved to the CNCF in 2016, it unlocked enterprise adoption that wouldn't have happened under Google's umbrella. MCP and Goose going neutral under AAIF could catalyze the same kind of adoption curve for agent frameworks — particularly in enterprises that require vendor-neutral procurement to approve new infrastructure.”