Memory

What it is

Memory is the context Claude carries from one chat to the next — how you work, what you’re working on, corrections you’ve made — so you don’t have to re-brief it every time. It focuses on work-related context.

You can see what Claude has stored, change it by telling Claude what to change in plain English, bring memory across from another AI service, or turn the whole thing down to just the parts you want.

Where to find it

On desktop

Settings → Memory.
Also reachable from Settings → Privacy → Memory preferences → Manage.

On the web

Settings → Memory, at claude.ai/new#settings/memory.

How to use it

Read and change what Claude remembers

The Memory pane lists what Claude has stored under You and Topics; the desktop inspection also reached an Areas section. Personal entries were excluded from the review.

At the foot of the pane there’s a box reading Tell Claude what to change or remove. Type an instruction — “forget that I use Postgres”, “remember I present to the board monthly” — and press Submit memory change. There’s no per-row edit form; you tell Claude and it rewrites the entry.

Choose how much Claude remembers

Three switches sit above the list:

Search and reference chats — lets Claude look for relevant details in past chats.
Generate memory from chats — lets Claude write new memory from what you talk about.
Include sensitive topics in memory — lets Claude save details about sensitive topics such as health conditions or religious beliefs.

Import memory from another AI service

In your other AI service, run the export prompt Anthropic provides in the import and export article — it asks that service to list every memory it holds about you.
Copy the result.
In Claude, go to Settings → Memory → Start import.
Paste into the box and choose Add to memory.

Export it

Open Settings → Memory and copy the contents into a local file. Or ask Claude in any chat: “Write out your memories of me verbatim, exactly as they appear in your memory.”

Good to know

Imports are lossy by design. Claude’s memory is built around work-related context, so personal details unrelated to work may not survive an import. Add the ones that matter by hand afterwards.
Sensitive-topic memory is optional. Include sensitive topics in memory can enable topics such as health information. Anthropic says government ID numbers, criminal history, financial account numbers and immigration status are never saved to memory, regardless of that switch.
Trim the export prompt first if you’d rather not carry passwords, health or finance details across from your other service.
Plans. Memory imports are available on Free, Pro, Max and Team plans, on the web and on Claude Desktop.
Memory is a per-chat toggle too. The composer’s + menu has a Memory checkbox, so you can run a single conversation without it.

What’s changed recently

2026-10-07 — The three memory switches and Start import are present on both platforms; desktop also shows an Areas heading. Personal memory content and editing were not inspected.
2026-09-13 — Memory has moved out of Settings → Capabilities into its own Settings → Memory pane on both platforms. The pane now shows stored memory grouped under You and Topics, adds three switches (Search and reference chats, Generate memory from chats, Include sensitive topics in memory), and is edited in natural language via a Tell Claude what to change or remove box rather than a per-entry form. Anthropic’s help centre calls this the new memory experience and the Capabilities location the legacy one. On desktop, Settings → Privacy also gained a Memory preferences → Manage shortcut into the same pane.
2026-04-18 — Page created from official help-center article.