claude
memory
the 3-layer setup
Claude remembers you now — across chat and Cowork. Most people never turn it on, so they explain themselves from scratch every time. Turn it on once. It works in three layers.
What it knows about you
Turn it on and Claude keeps one memory about you — the same memory in chat and in Cowork. One brain, everywhere.
- (1)Open Settings › MemoryIn Claude, open Settings, then Memory. Some accounts show it under Settings › Capabilities.
- (2)Turn on Generate memory from chatsNow it remembers across chat and Cowork — one shared memory.
- (3)Click View and edit your memoryRead your Topics — the things it has picked up about you.
Bring your ChatGPT memory
Optional. If you've spent months in ChatGPT, move that memory over in one paste. Starting fresh? Skip it.
- (1)Copy the prompt below
- (2)Paste it into ChatGPTIt prints everything it remembers in one code block.
- (3)In Claude: Settings › Memory › Start import
- (4)Paste the code block → Add to memoryIt's experimental, so read it before you trust it.
Memory per project
Every project gets its own memory too — sitting on top of what Claude already knows about you.
- (1)Every project has its own memoryAnd its own running summary of the work so far.
- (2)System memory = you. Project memory = the workClaude keeps the two apart on purpose.
- (3)Keep them separatePersonal facts stay in system memory. Project facts stay in the project.
Keeping it clean
View and edit your memory shows your Topics — what Claude thinks you are.
Edit or delete any Topic that's wrong or out of date.
Keeps everything it has, and stops writing anything new.
Wipes everything — including your project memories.
Set it up once, and every chat starts out already knowing you — no more starting from zero.
Compass rose: Catalan Atlas, 1375 · public domain. Engravings: public domain.
Get the next guide
Experiments like this one — thinking and working better with AI. Straight to your inbox, free.