Vovy reads your repo and writes a CLAUDE.md with your stack, commands, folder map and hard rules, so every Claude Code session starts with context.
Write a CLAUDE.md so Claude Code knows how my app works and what not to touch
How it works
- Scan the repo: Vovy reads package.json, your folder layout and config files to learn the framework, database and how the app is started.
- Ask about your rules: Vovy asks a few questions: what should never be edited, how you like commits, which commands to run before shipping.
- Draft the CLAUDE.md: Vovy writes a short file with sections for stack, commands, key folders and do-not rules. Claude Code reads it automatically at the start of every session.
- Review it with you: Vovy shows the draft as a card and trims anything vague. Short, specific rules work better than long essays.
- Test it in a session: Vovy starts Claude Code and asks how to run the app, confirming it answers from your new file.
What you provide
- Your project folder
- Your rules and preferences
What you get
- A concise CLAUDE.md in your repo
- Commands and folders documented
- Do-not-touch rules written down
- A tested Claude Code session
FAQ
Does Cursor read CLAUDE.md too?
Cursor uses its own rules files and also reads AGENTS.md. Vovy can create a matching version for Cursor if you use both.
How long should it be?
Short. A page or two of specific instructions beats a long document, because it is loaded into every session.
Should it go in git?
Yes, so collaborators and every AI session share the same rules. Keep personal notes in CLAUDE.local.md instead.
Related tasks
All tasks