AGENTS.md reaches 60,000+ repo adoptions across major AI coding tools

AGENTS.md, the open standard for giving AI coding agents project-specific guidance, has now been adopted by over 60,000 open-source projects and most major agent frameworks.

The standard works like a README file, but aimed at AI agents instead of human developers. You drop a markdown file in your repository root that describes your project's conventions, preferred tools, testing commands, deployment patterns, and anything else an AI agent should know before modifying code. The format is intentionally simple: plain markdown with no special syntax or schema to learn.

Support is now built into GitHub Copilot, Cursor, Codex, Devin, Factory, Gemini CLI, Jules, VS Code, and others. When an AI coding agent encounters an AGENTS.md file, it reads the guidance before making changes, which makes its behavior more predictable and aligned with the project's actual workflow. This means fewer corrections, less cleanup, and more useful first-pass code generation.

OpenAI originally released the standard in mid-2025, and it was donated to the Agentic AI Foundation under the Linux Foundation in December 2025. The vendor-neutral governance ensures no single company controls the format, which has been important for adoption across tools from competing companies like OpenAI, Anthropic, Google, and Microsoft.

The 60,000 adoption number is significant because it represents organic, bottom-up adoption. Developers are choosing to add AGENTS.md files because the agents actually work better when the guidance is there. It is not a theoretical standard waiting for adoption - it is a practical tool that developers add after noticing their AI agents make fewer mistakes with project context.

For anyone already using a project-specific file like CLAUDE.md for Claude Code, AGENTS.md serves a similar purpose but works across multiple tools. The two can coexist in the same repo: CLAUDE.md for Claude-specific settings (like allowed tools, security rules, and memory configuration) and AGENTS.md for universal agent guidance (like coding conventions, test commands, and deployment procedures).

Getting started is simple: create a file called AGENTS.md in your repository root, describe your project's conventions in plain language, and commit it. Include sections for the tech stack, testing commands, file structure, naming conventions, and any common mistakes agents should avoid. The AI agents that support the standard will pick it up automatically on the next interaction. No registration, configuration, or dependencies required.