AGENTS.md
AGENTS.md is an open, simple format for providing project-specific instructions to AI coding agents. It serves as a dedicated, predictable place to define context such as build commands, testing procedures, code style conventions, and environment setup—complementing human-focused READMEs. The format is designed to be a single file placed in the project root, read by a wide range of AI coding tools and platforms. The specification has been adopted by over 60,000 open-source projects on GitHub, demonstrating strong community endorsement. It works across numerous AI coding agents including VS Code, GitHub Copilot, Cursor, Devin, Aider, Windsurf, and many others, making it a de facto standard for agent instructions. AGENTS.md is intentionally kept minimal and agent-focused, avoiding clutter that might overwhelm human contributors. By separating agent instructions from documentation for people, it enables maintainers to provide precise, actionable guidance to coding agents without compromising the readability of existing project documentation.
- Category
- AI Agent
- Pricing
- Free
Problem solved
Coding agents often lack consistent, project-specific guidance, leading to incorrect or non-conforming code. AGENTS.md solves this by providing a standardized file that both open-source and closed-source projects can use to convey instructions to AI agents, ensuring they follow build steps, testing workflows, and style rules appropriately.
Use cases
- Define build and test commands for AI coding agents
- Specify code style guidelines (e.g., linting rules) for generated code
- Provide environment setup steps to ensure agent reproducibility
- Share project conventions with multiple AI tools via a single file
- Document directory structure and monorepo navigation for agents
- Instruct agents on CI/CD pipeline integration and deployment steps
- Enable agent-friendly contributor onboarding by including custom instructions
Audience
- Developers
- ML Engineers
- Data Scientists
- Researchers
Pros
- Simple, open, and tool-agnostic format
- Widely adopted by over 60k open-source projects
- Compatible with many leading AI coding agents
- Complements README, reduces clutter
- Easy to adopt—just add a file
Cons
- No enforced validation or syntax checks
- Limited to static instructions; cannot include dynamic logic
- Relies on agent developers to support the standard
- May require duplication if agents need different formats
- Can become outdated without manual updates
Official website: https://agents.md/
Canonical LambdaBase page: https://www.lambdabase.com/tools/agents-md