OpenAI, Anthropic, and Block launch Agentic AI Foundation to set global standards for AI agents

OpenAI is taking a major step in shaping how AI agents interact across tools and platforms. The company is co-founding the Agentic AI Foundation (AAIF) under the Linux Foundation alongside Anthropic and Block, with support from Google, Microsoft, AWS, Bloomberg, and Cloudflare. The goal is straightforward: build open, interoperable standards for agentic AI before the ecosystem splinters into incompatible approaches.

As part of this effort, OpenAI is contributing AGENTS.md, a lightweight Markdown format that provides AI agents with a clear, predictable way to read project-specific instructions.

“OpenAI is co-founding the Agentic AI Foundation (AAIF) under the Linux Foundation alongside Anthropic and Block to support open, interoperable standards for agentic AI. We’re also donating AGENTS .md to help establish open standards that enable safe, reliable agents across tools, repositories, and ecosystems,” the maker of ChatGPT said in its announcement on X.

OpenAI, Anthropic, and Block launch Agentic AI Foundation under the Linux Foundation to set global AI agent standards

Developers are already leaning heavily on AI agents for coding assistance, workflow automation, and support tasks. But 2025 marks a shift. These systems are moving from experimental demos to real applications, where they carry real responsibility. That shift raises the stakes. Without shared conventions, different tools risk drifting apart, making agents harder to trust, maintain, and transfer between environments.

This is the moment AAIF aims to address. The foundation offers neutral stewardship—something developers have been asking for as agentic systems grow more complex and more embedded in everyday work. The idea is simple: the industry needs a shared playbook before AI agents become too widespread to align.

OpenAI points to the past year as proof that open infrastructure can gain adoption when it solves real problems. The company released its Agents SDK, the Apps SDK, the Agentic Commerce Protocol, and open-source work, including the gpt-oss models and the Codex CLI. That CLI alone has helped merge more than two million public pull requests on GitHub, giving OpenAI a front-row seat to how developers actually build with agents.

A New Standard for AI Agents: OpenAI and Anthropic Co-Found the Agentic AI Foundation

The company has also been a core contributor to the Model Context Protocol (MCP), which now supports connectors and apps inside ChatGPT. More recently, OpenAI, Anthropic, and MCP-UI expanded the Apps SDK to the entire MCP community through MCP Apps, making it easier for developers to build interactive agent-driven tools.

These efforts set the stage for AAIF by showing that shared standards can move beyond theory and into daily use.

AGENTS.md is a good example of that shift. It began as a practical house rule inside OpenAI: a place where Codex could reliably find instructions, coding conventions, build steps, and testing details. By placing the file next to README.md, developers gained a predictable place to define how an agent should behave inside a codebase.

What started as a convenience evolved into a convention. Since its debut in August 2025, more than 60,000 open-source projects and agent frameworks—from Amp and Codex to Cursor, Devin, Factory, Gemini CLI, GitHub Copilot, Jules, and VS Code—have adopted it. That amount of agreement in a short time signals a broader appetite for shared, neutral formats. Donating AGENTS.md to AAIF ensures no single company sets its direction as adoption grows.

Each co-founder is contributing a core project to the foundation: OpenAI with AGENTS.md, Anthropic with the Model Context Protocol, and Block with goose. Together, these projects give AAIF a strong starting set of conventions that already have traction across the industry.

AAIF’s role is to provide a stable home where these standards can develop in the open with long-term stewardship. The Linux Foundation will oversee the effort, bringing decades of experience supporting widely adopted technologies such as the Linux Kernel, Kubernetes, Node.js, and PyTorch. Its involvement signals that the industry is aiming for durability, not a short-lived working group.

OpenAI says the intention is to serve developers, enterprises, and the broader open-source community by creating a shared base for building agentic systems. With founding support from OpenAI, Block, Anthropic, Google, Microsoft, Amazon, Bloomberg, and Cloudflare, the foundation is grounded in a collective commitment to avoid the pitfalls of fragmentation as agentic AI moves into production.

The AAIF is now inviting contributors—tool builders, researchers, companies, and individual developers—to help shape what comes next. The hope is that a neutral home, backed by real adoption and shared investment, can keep agentic AI on a path that benefits everyone building in the space.