Skip to main content
VibeStacks

Loading the search index…

EntryTop pick

AGENTS.md

A README for AI coding agents. One Markdown file in the repository root that almost every tool reads.

Last reviewed
Website
agents.md

Good for

  • Every project that uses AI coding tools, without exception
  • Teams where some people use Cursor and others use a CLI agent, and everyone needs one shared set of instructions
  • Open-source projects, so contributors' agents follow the same conventions

Not for

  • Rules that apply only to certain folders or file types (use Rules)
  • Long procedures needed only for specific tasks (use Skills)

AGENTS.md is an open format: put a Markdown file in the repository root that tells an agent what it needs to know to work on the project. README is for humans; AGENTS.md is for agents. It emerged from tools including OpenAI Codex, Cursor, Google Jules, Amp, and Factory, is now stewarded by the Agentic AI Foundation under the Linux Foundation, and is used by more than 60,000 open-source projects.

Why it is our top pick

  • Works across tools. Cursor, Codex, Pi, and others read it directly, so you do not rewrite it when you switch. Claude Code reads CLAUDE.md by default; make CLAUDE.md a one-line pointer to AGENTS.md, or symlink it, to keep a single source of truth.
  • No format overhead. It is plain Markdown with no required fields; use whatever headings you like.
  • It can be layered. In large repositories, subfolders can have their own AGENTS.md, and the one closest to the edited file takes precedence.

What to write

  1. Commands. How to install, run, test, and check. Agents need this most and guess it wrong most often.
  2. Conventions. Code style, what each folder is for, naming, commit message format.
  3. Boundaries. What needs permission first (committing, deleting, changing public interfaces) and which files are off-limits.
  4. Verification. Which checks must pass after a change.

What not to write

  • Background stories and marketing copy meant for human readers.
  • General knowledge the model already has ("write clean code").
  • Long procedures needed for only a few tasks; turn those into Skills.
Souvenir

Take it home to your AI

Drop this AGENTS.md into your project, or paste it straight to your AI coding assistant.

AGENTS.md
# AGENTS.md

## Overview
- <one-sentence project description>
- Stack: <e.g. Vite + React + TypeScript, pnpm>
- Folders:
  - `src/features/` modules split by business feature
  - `src/components/` UI components shared across features
  - `src/lib/` framework-agnostic utilities

## Commands
- Install: `pnpm install`
- Dev: `pnpm dev`
- Unit tests: `pnpm test`
- All checks: `pnpm check`

## Conventions
- TypeScript strict mode, no `any`; validate external input at the boundary.
- Component files use PascalCase; other files use kebab-case.
- Commit messages follow Conventional Commits.

## Boundaries
- Ask before committing, pushing, deleting files, or changing public interfaces.
- Do not read or print secrets from `.env*`.
- Do not hand-edit generated code or lockfiles.

## Definition of done
- `pnpm check` passes; new behavior has tests; when fixing a bug, first add a test that reproduces it.
  • Article

    The pieces of context you give AI

    AGENTS.md sets project rules, Rules handle local scope, MCP adds external capabilities, and Skills hold know-how loaded on demand.

  • EntryAlternative

    Rules

    Tool-specific rules that apply only to certain folders or file types, covering local conventions that do not belong in AGENTS.md.

  • EntryTop pick

    Pi

    A minimal, self-extending terminal coding agent that gives you primitives and leaves the workflow to you.

  • EntryTop pick

    Cursor

    An AI IDE built on VS Code. Completion, chat, the agent, and your code live in one window, and every change arrives as a diff.