AGENTS.md
Open format that guides AI coding agentsA simple AGENTS.md file that tells agents how to work
Highlights
What you can do with it
- Guide an AI agent to write and test code in an open-source project
- Give sub-projects in a monorepo their own instructions
- Offer agents a standard input format for better project fit
How to use it
- 1 Create an AGENTS.md file in the root of your repository
- 2 Add sections such as overview, build and test commands, and style
- 3 Include commit, PR, security, and deployment notes as needed
- 4 For a monorepo, add a nested AGENTS.md in each sub-project
Who it is for
Pros
- Simple, open, and free to adopt
- Keeps agent instructions separate from the human README
- Scales to large monorepos with nested files
- Broad agent and tool compatibility
Limitations
- It is a format, not a tool, so results depend on the agent
- Free and open now, but stewardship and terms can evolve
- Needs upkeep as build and test commands change
- Only as good as the instructions you write
What AGENTS.md is
AGENTS.md is a simple open format used to guide coding agents. You add an AGENTS.md file to your repository with a project overview, build and test commands, code style rules, and submission guidelines, so an AI coding agent gets clear, predictable instructions instead of guessing. It complements the README by keeping agent-specific detail in one place, and for large single-repository projects you can nest AGENTS.md files so each sub-project gets its own guidance while the agent reads the nearest one. It is compatible with many AI coding agents and tools, and it is the result of collaboration across the AI software ecosystem, including OpenAI and Google, hosted by the Agentic AI Foundation under the Linux Foundation. It is free and used by over 60k open-source projects.
What you can do with it
- Give agents exact install, build, and test commands
- Set code style and commit and PR rules for consistent output
- Guide agents in large monorepos with nested files per package
- Provide a standard input format for many coding agents and tools
Who it is for
- Software developers who work alongside coding agents
- Open-source maintainers coordinating many contributions
- AI coding agent developers who need a standard project format
- Teams that want predictable agent behavior in a repo
What to watch out for
AGENTS.md is a format, not a tool, so the results depend on the agent that reads it and how well you write the instructions, and a vague file gives vague behavior. It needs upkeep as build and test commands change, or the agent will follow stale steps and waste time. It is free and open now, but stewardship and terms can evolve, so check the current project before you standardize on it. Keep sensitive details out of the file if your repository is public, since anyone can read what you put there.
FAQ
What is AGENTS.md?
It is a simple open format, a file in your repo, that gives coding agents clear instructions.
Is it free?
Yes, it is free and is hosted by the Agentic AI Foundation under the Linux Foundation.
Does it work in a monorepo?
Yes, you can use nested AGENTS.md files so each sub-project gets its own guidance.
Why it is free
What the official page says： 免费
Free limits, trial length and features are decided by the official site and can change at any time.
Product preview
Screenshot archived from the listing page, for preview only.
