AGENTS.md Generator
Generate project rule files for Cursor, Claude Code and other coding agents.
Done calculating? Check these tools
Cursor
Cursor is an AI-powered code editor built for pair programming with AI. It…
Claude Code
Claude Code is Anthropic's CLI tool for AI-powered coding. It brings Claude's…
Windsurf
Windsurf is an AI-first IDE by Codeium with powerful features like Cascade AI…
Related categories
About AGENTS.md Generator
Cursor, Claude Code, Codex and Gemini CLI all read AGENTS.md — the file that decides how well AI coding agents behave in your repo. Pick a project type (Next.js frontend / Python RAG / Playwright scraper / n8n workflows) and the generator pre-fills commands, architecture notes, conventions and red lines from real engineering practice for that stack; tweak to taste and copy the finished AGENTS.md. Also covers the CLAUDE.md symlink compatibility trick and exports .cursor/rules/*.mdc (the legacy .cursorrules is deprecated). Everything generates locally in your browser.
FAQ
- What is AGENTS.md, and how does it relate to CLAUDE.md and .cursorrules?
- AGENTS.md is an open convention: the file at the repo root is auto-read by mainstream AI coding tools as the project's onboarding brief for agents. It's vendor-neutral — Cursor, Claude Code, Codex and Gemini CLI all support it — so maintaining this one file covers everything. The legacy .cursorrules is deprecated by Cursor (rules move to .cursor/rules/*.mdc), and recent Claude Code reads AGENTS.md natively; `ln -s AGENTS.md CLAUDE.md` covers older setups in one line.
- Are the generated templates any good?
- The templates are not fill-in-the-blanks: each of the four types is written from real engineering practice — the RAG template demands eval deltas for retrieval changes, the scraper template defaults to conservative rate limits and forbids network access in tests, the frontend template guards the server/client boundary and lazy-loading, the n8n template insists on exporting workflow JSON back to Git. Commands come pre-filled with each stack's standards; you only adjust what differs.
- What actually belongs in an AGENTS.md?
- The four categories agents get wrong most: ① how to verify (install/test/lint/build commands); ② where code lives (directory responsibilities); ③ this project's unique workflow (eval gates, export procedures); ④ absolute red lines (credentials, generated files). Generic "write clean code" advice carries no signal — rules concrete enough to name commands and paths are what agents can act on.