Claude Code Basics — Give Claude a Memory (CLAUDE.md)

Claude Code Basics · stop repeating yourself

Give Claude a memory

Start a new session and Claude forgets everything — your stack, your commands, your style. A CLAUDE.md file is its permanent memory for this project. Write it once; it's there every time.

📄

One file in your project — Claude reads CLAUDE.md automatically, at the start of every session.

⚡

Create it in one step — run /init and Claude scans your project and writes a starter for you.

✍️

Yours to shape — edit it, or just say "add this to CLAUDE.md" and Claude does it.

No more re-explaining your project on every fresh chat.

What to put in it

A good memory has five things

1

How to run it

The commands to install, start, and preview the project.

2

How to test it

The test command — so Claude can check its own work.

3

Your style

Naming, formatting, the conventions you want followed.

4

The layout

What lives where — the project's structure, in a few lines.

5

Gotchas

The "always do this / never touch that" rules unique to your project.

A well-organized memory file with clear sections

Keep it tight — about a page. A short, sharp memory beats a long rambling one.

Two kinds of memory

Team memory vs your memory

👥

Project memory — CLAUDE.md in your project. Shared with your team through git. The build steps & rules everyone follows.

🙋

Your personal memory — ~/.claude/CLAUDE.md. Just you, across every project. Your shortcuts and preferences.

👀

See it all — run /memory to view and edit everything Claude is remembering.

Keep secrets out of the shared file — those belong in your personal memory or a .gitignored CLAUDE.local.md.

Write it once. Claude remembers.