AGENTS.md generator (free, no signup)
Answer a few questions about your project and get a ready-to-commit AGENTS.md your coding agents can follow. Fill in the setup command, build and test steps, conventions, and paths to leave alone, then copy the Markdown. Runs in your browser, nothing you type is sent anywhere.
Generator
Build your AGENTS.md
Optional sections
# my-app A Node.js REST API with a SQLite database. ## Setup ```bash npm install ``` ## Build and test ```bash # Build npm run build # Test npm test # Typecheck npm run typecheck ``` ## Conventions - Prefer named exports over default exports - TypeScript strict mode is on; avoid `any` - No raw SQL outside src/server/db.ts ## Do not edit `dist/`, `node_modules/`, `data/`
Add the repository as owner/repo, for example octocat/hello-world.
Why it matters
What goes in an AGENTS.md
A coding agent starts each session with no memory of the last one. Without a context file, it has to infer the setup steps, guess the naming conventions, and sometimes overwrite files it should not touch. An AGENTS.md fixes that in one commit. The agent reads it before writing a single line of code.
- Commands first. Setup, build, test, and lint are the highest-value lines. An agent that knows how to run tests can check its own work without being asked.
- Conventions the code does not show. If the codebase has a naming pattern, a banned dependency, or a commit message format, write it down. Agents infer style from what they see; they need to be told about things they cannot see.
- Paths to leave alone. Generated files, data directories, and vendor bundles do not belong in a diff an agent authors. List them so the agent stops before it writes there.
- Keep it short. Every token in AGENTS.md is a token not spent on the actual task. A minimum useful file is under 30 lines. Omit anything the code already makes obvious.
Claude Code reads both AGENTS.md and CLAUDE.md; OpenAI Codex reads AGENTS.md; Cursor reads .cursorrules. AGENTS.md is the cross-agent standard. If your repo already has a CLAUDE.md, the two can be identical or you can symlink one to the other. The Quell agents page covers the other half: once the agent knows how the repo works, Quell feeds it agent-readable bug reports to act on through a REST API, CLI, or MCP server.
FAQ
AGENTS.md questions
What is an AGENTS.md file?
An AGENTS.md file is a plain text context document that coding agents read at the start of every session to understand a repository. It contains the setup commands, build and test steps, file paths that should not be touched, and conventions the codebase follows. Claude Code reads AGENTS.md and CLAUDE.md; OpenAI Codex reads AGENTS.md; Cursor reads .cursorrules. Writing one once saves the agent from guessing the same things every session.
What should I put in an AGENTS.md?
The minimum useful set is the setup command, the build command, the test command, and the paths the agent must not write to. Beyond that, add conventions the codebase follows that a reader would not infer from the code alone: a naming pattern, a banned dependency, a commit message format. Omit anything the code already makes obvious.
What is the difference between AGENTS.md and CLAUDE.md?
Claude Code reads both AGENTS.md and CLAUDE.md when it starts a session. The difference is audience: AGENTS.md is the cross-agent standard, read by Claude Code, Codex, Cursor, and other agent runtimes; CLAUDE.md is specific to Claude Code and can carry Claude Code specific instructions. If you write one file and want all agents to benefit, use AGENTS.md. The two files can be identical, or one can be a symlink to the other.
Where does AGENTS.md go in a monorepo?
Put a root-level AGENTS.md with the repo-wide conventions and commands. Add a second AGENTS.md in each package or app directory that has its own build target, test runner, or set of conventions. A coding agent in a monorepo reads both the root file and the nearest package-level file, so the package file can stay focused on what is different without repeating the whole repo context.
Is this AGENTS.md generator free?
Yes. There is no signup and no limit. Fill in the fields and copy the Markdown. It runs in your browser; nothing you type is sent anywhere.