AGENTS.md
AGENTS.md is an open standard file format that provides context and instructions to AI coding agents working on software projects. Like a README for agents, an AGENTS.md file lives in the project repository and tells AI agents how to build, test, and contribute code — including coding standards, build commands, testing procedures, and development conventions. Supported by over 60,000 open-source projects and major platforms including OpenAI Codex, Google Jules, Cursor, Devin, Windsurf, GitHub Copilot, and goose. Stewarded by the Agentic AI Foundation under the Linux Foundation.
APIs
AGENTS.md Specification
The AGENTS.md specification defines a standard Markdown file format for providing project context, build instructions, coding standards, and testing procedures to AI coding agen...
Features
Provides AI agents with structured information about the project including its purpose, architecture, and key concepts.
Documents the exact commands needed to build, test, lint, and deploy the project so AI agents can execute them correctly.
Specifies coding conventions, style guides, naming patterns, and best practices that AI agents should follow when generating code.
Supports nested AGENTS.md files in monorepo subdirectories, allowing different projects to have tailored agent instructions with closest-file-wins precedence.
The format uses standard Markdown with no mandatory fields, giving teams flexibility to include only the context their agents need.
Supported by 60+ AI coding tools and editors including OpenAI Codex, Google Jules, Cursor, Devin, GitHub Copilot, Windsurf, and goose.
Use Cases
Provide AI coding agents with project conventions and standards so their generated code passes review without needing manual style corrections.
Enable AI agents to independently implement features by giving them the context needed to set up the development environment, run tests, and validate their work.
Accelerate AI agent productivity on legacy codebases by documenting the context that would normally come from exploring the project.
Ensure all AI agents working on a project operate from the same context, reducing inconsistencies when multiple agents collaborate.
Document security considerations and sensitive file locations so AI agents avoid introducing vulnerabilities or accidentally exposing credentials.
Integrations
Native AGENTS.md support for AI coding tasks via OpenAI's coding agent.
AGENTS.md context ingestion in Google's AI coding agent for autonomous development tasks.
AGENTS.md file discovery and context loading in Cursor's AI-assisted editor.
AGENTS.md support in Cognition's autonomous coding agent Devin.
AGENTS.md integration in GitHub Copilot for project-aware AI code suggestions.
AGENTS.md context loading in Block's open-source goose AI agent, now under the Agentic AI Foundation.
AGENTS.md support in Windsurf's AI-powered development environment.
AGENTS.md discovery in JetBrains' AI coding assistant Junie.