Guides / Make Claude remember
How to make Claude remember across chats: 3 ways, compared
There are three ways. Claude's memory saves notes on its own as you chat. A Project gives a group of chats shared instructions and files. A CLAUDE.md file is a document you write, which Claude reads at the start of every session. For anything you rely on every day, such as clients, prices and rules, the file is the most dependable, because you decide exactly what is in it.
1. Turn on Claude's memory
With memory on, Claude saves short notes on topics while you chat, and uses them in later conversations. You can also tell it directly: "Remember that I bill by the day, not by the hour."
- Good for: small preferences you would not think to write down.
- Good to know: each Project has its own separate memory, and incognito chats are not remembered. On paid plans Claude can also search your earlier chats when you ask it to.
- Limit: Claude chooses what to save and how to word it. You can review and delete the notes, but you are not writing them.
2. Use a Project for each area of work
A Project groups chats that belong together, for example one per client. You give it instructions ("Lumen's brand voice is calm and plain") and files (the brand guide, the last proposal), and every chat in that Project starts with them.
- Good for: keeping one client's material together and out of the others.
- Limit: knowledge does not cross between Projects, and the instructions live inside the Claude app rather than in your own files.
3. Write a CLAUDE.md file
CLAUDE.md is a plain text file in the folder you work in. Claude Code reads it at the start of every session, including in the Code tab of the Claude desktop app. It is the method Anthropic documents for giving Claude lasting instructions.
- Good for: everything you depend on: who you are, how you work, your rules, and pointers to more detailed files.
- Strength: you write it, you can read it, and it sits on your computer. It survives any chat, and you can back it up or move it.
- Limit: it is only as good as you keep it. Claude's docs recommend under about 200 lines per file, specific wording, and no contradictions.
Our guide CLAUDE.md explained for people who don't code walks through one with a template.
Side by side
| Memory | Project | CLAUDE.md | |
|---|---|---|---|
| Who writes it | Claude | You | You |
| Where it lives | In the Claude app | In the Claude app | In a folder on your computer |
| Scope | All chats, or one Project | The chats in that Project | Every session in that folder |
| Can you read all of it? | You can review the notes | Yes | Yes, in any editor |
| Best for | Small preferences | One client or area | The context you rely on daily |
Which one should you use?
They are not rivals, and you can use all three. A simple rule of thumb:
- Write down what you would hate to explain again (clients, prices, tone, rules) in a
CLAUDE.mdand the files it points to. - Leave small habits to memory. If a note it saved is wrong, delete it.
- Every few weeks, read your files. Move what is no longer true to an archive, and remove rules that contradict each other.
Why the written file matters most is explained in Why does ChatGPT or Claude forget?
Rather not set this up yourself?
Stignova sets up Claude with your own files, fills them in with you on a call, and proposes updates every week that you approve.
Join the early list