Universal Standard

AGENTS.md

AGENTS.md is an open standard for AI coding agent instructions. It provides a consistent format that works across multiple AI tools and IDEs.

The Universal Format

AGENTS.md is used by 60,000+ projects and is supported by multiple AI coding tools. When you create an AGENTS.md file, it works across Cursor, Claude, Copilot, and many other tools that support the standard.

Visit agents.md

Configuration File

AGENTS.md

A markdown file placed in your project root that any AGENTS.md-compatible tool can read.

Why Use AGENTS.md?

One File, Many Tools

Write your instructions once, use them across all compatible AI tools

Future-Proof

New AI tools will likely support AGENTS.md as it becomes the standard

Team Consistency

Everyone on the team gets the same AI behavior regardless of their preferred editor

Community Standard

Share and discover configurations that work everywhere

Example AGENTS.md

# AGENTS.md

## Project Overview
E-commerce platform with React frontend and Go backend.

## Tech Stack
- **Frontend**: React 18, TypeScript, Vite, Tailwind CSS
- **Backend**: Go 1.21, Gin, PostgreSQL
- **Infrastructure**: Docker, Kubernetes, AWS

## Architecture
```
/frontend    - React SPA
/backend     - Go API server
/deploy      - Kubernetes manifests
/docs        - API documentation
```

## Code Style

### TypeScript/React
- Use functional components with hooks
- Prefer named exports
- Use absolute imports (@/components/...)
- Comprehensive TypeScript types

### Go
- Follow standard Go conventions
- Use interfaces for dependencies
- Table-driven tests
- Structured logging with slog

## Git Workflow
- Conventional commits (feat:, fix:, docs:, etc.)
- Feature branches from main
- Required PR reviews
- CI must pass before merge

## Testing
- Frontend: Vitest + Testing Library
- Backend: Go testing + testify
- E2E: Playwright
- Minimum 80% coverage for new code

## Security
- No secrets in code
- Use environment variables
- Input validation on all endpoints
- SQL parameterized queries only

Compatible Tools

These tools read AGENTS.md by themselves:

✓OpenAI Codex
✓Cursor
✓Devin Desktop (formerly Windsurf)
✓Antigravity
✓Zed
✓Cline
✓Roo Code / Zoo Code
✓Augment Code
✓Kilo Code
✓Junie
✓Kiro
✓Goose
✓Warp
✓OpenCode
✓OpenHands
✓Crush
✓Firebender

Claude Code and Gemini CLI read their own file, so LynxPrompt writes a one-line CLAUDE.md or GEMINI.md that imports AGENTS.md with @AGENTS.md. Aider loads no rules file by itself, so LynxPrompt writes a .aider.conf.yml that lists AGENTS.md under read:. GitHub Copilot, Trae, Tabnine and Firebase Studio get a full copy of the rules in their own file.

Monorepo Support

AGENTS.md supports hierarchical configurations for monorepos and multi-package projects. Place AGENTS.md files at multiple levels:

my-monorepo/
├── AGENTS.md              # Root: shared rules for entire repo
├── packages/
│   ├── web/
│   │   └── AGENTS.md      # Web app specific rules
│   ├── api/
│   │   └── AGENTS.md      # API server specific rules
│   └── shared/
│       └── AGENTS.md      # Shared library rules
└── tools/
    └── AGENTS.md          # Build tools specific rules

Inheritance

AI tools read the closest AGENTS.md to the file being edited, while also considering parent-level rules for shared context.

Specificity

Package-level AGENTS.md can override or extend root rules with specific tech stacks, conventions, or boundaries.

Tip: Use the root AGENTS.md for project-wide rules (git workflow, security policies, team conventions) and subfolder AGENTS.md for package-specific rules (tech stack, testing frameworks, code style).

LynxPrompt + AGENTS.md

LynxPrompt generates AGENTS.md files by default when you use the wizard. You can also:

Pick several tools at once: the rules go in one AGENTS.md, and each tool gets only the file it needs to reach it
Create blueprints that include AGENTS.md
Use AI editing to customize your AGENTS.md
Generate separate configurations for each package in your monorepo

← Devin Desktop FAQ →