Free tool · Eddie’s Workshop
CLAUDE.md generator.
Answer a few questions and get a CLAUDE.md that tells Claude Code how your project works, so it stops rediscovering your commands every session.
Free · no sign-up · updated October 2026
What is CLAUDE.md?
CLAUDE.md is a Markdown file Claude Code loads into its context at the start of every session. It’s where you put what the agent can’t easily figure out by reading the code: the exact commands to run, your conventions, the folders that matter and the mistakes to avoid.
Put it at the root of your repo and commit it, so everyone on the team (and every agent) gets the same instructions. Running /init in Claude Code can draft one from your codebase; this generator gives you a clean starting point when you’d rather decide what goes in.
CLAUDE.md best practices
Keep it short. It’s loaded every session, so every line costs context. Aim for under 100 lines.
Write commands exactly. “Run tests” is vague; npm test -- --watch=false is not. Check each one by running it.
Say what not to do. Folders to leave alone and things that look safe but aren’t save more time than style rules.
Update it when the agent gets something wrong twice. That’s the signal it’s missing a line.
Made by Eddie
Stop checking on your agents.
Your CLAUDE.md keeps Claude Code on track. Eddie tells you the moment it needs you, right from your notch.
Questions.
Where does CLAUDE.md go?
At the root of your project (or in .claude/CLAUDE.md), committed to the repo. Claude Code also reads ~/.claude/CLAUDE.md for instructions that apply to all your projects, and CLAUDE.local.md for personal notes you keep out of git.
What’s the difference between CLAUDE.md and AGENTS.md?
CLAUDE.md is Claude Code’s file. AGENTS.md is an open format read by Codex, Cursor and other agents. To use both without two copies, keep your instructions in AGENTS.md and put @AGENTS.md at the top of CLAUDE.md: Claude Code imports the file, and Codex reads AGENTS.md directly.
How long should a CLAUDE.md be?
Short. It’s loaded into every session, so keep it to what the agent truly needs: commands, conventions, boundaries and gotchas. Under 100 lines is a good target.