CLAUDE.md generator
A CLAUDE.md is a short note you leave in your project telling the AI how things work here, which it reads before it changes anything. Answer the questions below, then copy the file straight off the page or have us email it to you. It takes about two minutes and you do not need to know how to code to fill it in.
# CLAUDE.md Notes for whoever works on this next, human or otherwise. Keep it short enough that it stays true. ## How to work in this project 1. Write code that reads like the code around it. Match the existing naming, comment density, and idioms rather than importing your own conventions. 2. Make the change that was asked for. Do not refactor surrounding code, add abstractions, or handle cases that cannot happen. If you think something else needs fixing, say so instead of doing it. 3. Run the test command before reporting that work is finished. If tests fail, say so and paste the output. Never describe unverified work as done. 4. Do not add a dependency without asking. Check what the project already has first, and prefer it even if it is slightly clumsier. 5. Ask first before anything hard to undo: deleting files, dropping or altering database tables, force pushing, rewriting git history, or touching production. 6. Never commit secrets, and never print an API key, token, password, or customer record into logs, terminal output, or an error message.
Where it goes: save it in your project’s main folder, the one with all your other files in it. Claude Code finds it by itself every time it starts. If you use Cursor instead, save the same text as a file called .cursorrules.
Why bother with this at all
Every time you start a new chat, the AI has forgotten everything. It does not remember that you tried something last week, or that one part of your project is fragile, or how you like things done. So it guesses, and it guesses the same wrong things over and over, and you correct it over and over.
A CLAUDE.md is just a note it reads first, every time, so it stops guessing. That is the whole idea. It is an ordinary text file you can open and edit yourself.
The box that matters most
Not the rules. The one labelled check it still works.
If the AI knows how to check its own work, it will: it makes the change, runs the check, sees it broke something, and fixes that before it ever tells you it is done. If it does not know how, it writes the code, tells you it is finished, and you find out later. That single line changes the experience more than every rule underneath it.
Second most useful is the last box, the surprising things. Everything else here the AI could eventually work out by reading your files. That your test version is secretly connected to the real thing is not something it can work out, and it is exactly the sort of surprise that turns a small mistake into a bad afternoon.
Why none of the rules shout
You may have seen these files written in capitals: CRITICAL: YOU MUST ALWAYS. Older AI models needed that to take an instruction seriously at all, and plenty of examples online still look like it.
Current models pay close attention to this file, so shouting now backfires: told it MUST do something, the AI starts doing it in situations where it should not. Calm, ordinary sentences work better, and they are easier for you to read back later.
Short beats complete
The way these files go wrong is not by being too small. It is by growing a new rule every time something annoys you, until there are thirty of them, half are out of date, and nobody remembers why. When a rule stops being true, delete it. A short file you trust beats a long one you do not.
If you have not set up an AI coding assistant yet, our beginner guide starts from nothing installed. If you want to go further, the memo on ECC covers the free setup that made this idea popular in the first place.
Questions people actually ask
What is a CLAUDE.md file, in plain English?
A short note you leave in your project that tells the AI how things work here. It is an ordinary text file you can open and edit yourself. The AI reads it before it changes anything, every time, so you stop having to explain the same things in every new chat.
Where exactly do I put it?
In your project's main folder, the one that has all your other files in it. Claude Code finds it on its own, with nothing to switch on. If you use Cursor instead, save the same text as a file called .cursorrules. Most other AI coding tools look for one of those two, and the wording works the same in all of them.
I do not know my commands. Can I still use this?
Yes. Pick the closest project type at the top and we fill in the commands that kind of project usually uses, then leave anything you are unsure about empty. A missing command is fine; a wrong one is worse than none, because the AI will trust it and get confused when it fails.
Do I need to know how to code?
No. Every box on this page can be answered in ordinary English, and the ones that need a command are pre-filled for you. The most valuable box is the last one, where you describe anything surprising about your project, and that needs no technical knowledge at all.
How long should the file be?
Short enough that it stays true. These files go wrong by growing a new rule every time something irritates you, until half of them are out of date and the AI is following a rule that stopped being right months ago. When something stops being true, delete it.
Is anything I type sent to you?
No. The whole thing runs inside your browser and nothing leaves the page. There is no account and no upload, so you can safely put your real commands and real details in. You can turn off your internet connection and it still works.
This tool is the short version of a memo. The free file that makes your AI code like a pro explains what is actually happening and why it works, with every claim sourced.
Want this built into your business?
We make these tools because we build this stuff every day. If you want it working inside your own company instead of sitting in a browser tab, that is literally our job.