Generate and safely maintain an AGENTS.md for any repo, the cross-tool standard that AI coding agents read.
One AGENTS.md works across Codex, Cursor, GitHub Copilot, Gemini CLI, Aider, Windsurf, Zed, Devin, and more. CLAUDE.md is pointed at it via symlink. agents-md writes a great one by reading your repo, and it is safe to re-run: it only updates the regions it generated and never clobbers anything you wrote by hand.
cd your-repo
npx @eugeniughelbur/agents-mdOr as a Claude Code skill: npx skills add eugeniughelbur/agents-md, then run /agents-md.
AGENTS.generated.md is written instead.| File | Read by | What it is |
|---|---|---|
| AGENTS.md | Codex, Cursor, Copilot, Gemini CLI, Aider, Windsurf, Zed, Devin | The cross-tool standard for agent instructions |
| CLAUDE.md | Claude Code | Tool-specific; symlink it to AGENTS.md |
| .cursorrules | Cursor | Tool-specific (Cursor also reads AGENTS.md) |
| .github/copilot-instructions.md | GitHub Copilot | Tool-specific (Copilot also reads AGENTS.md) |
| llms.txt | LLMs / AI search | For documentation sites, not repo agent instructions |
AGENTS.md is an open, cross-tool standard file that tells AI coding agents how to work in your repository: its build and test commands, code conventions, gotchas, and do-not-touch boundaries. It originated at OpenAI in 2025 and is now governed by the Linux Foundation's Agentic AI Foundation. Think of it as a README for agents.
Use AGENTS.md as the single source, then symlink CLAUDE.md to it. Most agents read AGENTS.md natively; Claude Code reads CLAUDE.md, so the symlink covers both. agents-md sets this up automatically.
Yes, it is safe. Re-runs refresh only the regions wrapped in HTML-comment markers; everything outside the markers is preserved. A handwritten AGENTS.md with no markers is never overwritten.
No. The CLI has zero dependencies, makes no network calls, and reads only repo metadata (and .env.example, never real secret values).