What Claude Remembers About You (2026 Guide)

At a glance

What this is: A walkthrough of Claude’s memory: where to find what it has saved, how to edit or delete it, and which settings are already on without you choosing them.

Who it is for: Anyone using Claude on a personal plan, and anyone using it at work who assumes their employer cannot see their private-looking chats.

The one thing to know: Deleting a conversation does not delete what Claude learned from it. Those are two separate things in two separate places.

Claude keeps notes on you now. Not a transcript, but a running set of facts it has decided are worth carrying between conversations: what you do, who you work with, how you like things written. On Free, Pro and Max plans this is on unless you turned it off. Most people have never opened the settings page where it lives.

This is worth ten minutes of your attention. Two of the defaults surprise people: memory is already on, and incognito chats are not private from your employer.

Table of Contents

Where do you see what Claude has saved?

Go to Settings > Memory. What Claude has saved from your ordinary chats is listed under Topics. Open any one to read it, then use the edit icon to change the wording or Delete to remove it. If you are on a Team or Enterprise account and the panel is empty or missing, memory has not been switched on for your organisation, and the rest of this section will not apply until it is.

Do this today even if you change nothing else. Claude writes these entries by itself as you chat, so it occasionally records something you said once, in passing, as though it were a standing fact about you. Reading the list is the bulk of the ten minutes.

You can also tell it to save something deliberately. Say remember this in a chat and it will save it, unless memory is paused or the fact falls into one of the categories below.

What is already switched on before you touch anything?

The defaults differ by plan, and not in the direction you would guess from the price. On a personal plan memory is already on and you have to switch it off. On a work plan it stays off until someone switches it on. The other settings do not follow that rule, so read the table.

Setting	Free, Pro, Max	Team, Enterprise
Memory (Claude saves what it learns)	On by default	Off until an owner enables it for the organisation
Search past chats	Not on Free. On by default for Pro and Max	On by default, unless your company runs its own encryption setup. Ask IT if you are unsure
Sensitive topics in memory	Off until you opt in	Off until you opt in
Incognito chats	Available, and excluded from memory	Available, but still covered by company data retention and exports

The sensitive-topics setting, and what turning it on changes

By default Claude refuses to save anything it reads as personal: health, race, ethnicity, religious belief, politics, gender identity and similar. If you would rather not re-explain a medical condition every time, you can turn on Include sensitive topics in memory in the same settings panel.

Three details matter if you do. It only works forwards, so nothing from before you flipped it gets saved. Each time it saves something in this category a notice appears above the message box. And if you later switch it back off, Claude deletes the sensitive entries it had already stored. Know that before you flip it: switching off is not a pause, it is a delete, and anything you told it about a health condition goes with it.

Claude is built not to save a further set of details at all: government ID numbers, criminal history, financial account numbers and immigration status. It tells you when it is refusing on those grounds. That is not a reason to type them in and see.

This is the one to remember. Anthropic states this plainly: when a conversation expires or is deleted, the memory entries generated from it are not removed. If you want something forgotten, delete the entry in Settings > Memory. Clearing your chat history does not do it, and all memory data is included in account data exports.

How do you turn it off without losing everything?

Switching memory off gives you two choices, and they are not the same:

Pause. Claude keeps what it already has but stops using it and stops adding to it. Anything you discuss while paused stays unsaved. It does not get picked up when you switch back on. Reversible.
Reset. Permanently deletes every memory, including the separate memories held inside each project. Anthropic is explicit that this cannot be undone.

Pause is the one you want if you are unsure. Reset is for starting clean and accepting the loss.

Projects each get their own memory

If you use projects, each one holds a separate memory and its own summary. Client work stays out of your personal planning, and one project does not leak into another. Searching past chats respects the same boundary: inside a project, Claude searches that project only.

Because those stores are separate, clearing your main Topics list does not touch them. If you want a project forgotten, deal with it inside that project. Choosing Reset in the memory settings is the one action that clears everything, projects included, and Anthropic is explicit that it cannot be undone.

The word memory makes people imagine something communal. It is not. What Claude knows follows you between the app, the web and Cowork. It does not travel between projects, and it never travels between people. Your colleague cannot see what Claude remembers about you.

One memory across chat and Cowork, with a catch

What Claude picks up in ordinary chat is available when you hand it a job in Claude Cowork, and what happens in a Cowork task comes back the other way. Ask it to draft a note to your manager and it already knows who that is.

The catch: this only works when Cowork runs in the cloud. Cowork sessions running locally on your own computer do not use memory at all. That is usually the explanation when Cowork seems to have forgotten what chat knows.

Incognito is private from Claude, not from your employer

The ghost icon on a new chat, top right on desktop, opens an incognito conversation. It is not saved to your history and Claude will not draw on it later or save memories from it. On a personal account that is exactly what it sounds like.

On a Team or Enterprise account it is narrower than the name suggests. Anthropic states that incognito chats are included in standard data exports and follow your organisation’s retention policies. The mode stops Claude remembering, but the conversation is still visible to the company paying for the account. Treat an incognito chat on a work account the way you would treat anything typed on a company laptop.

If Claude forgot something in the changeover, there is a deadline

Anthropic replaced the old memory system and migrated everyone across. If Claude seems to have lost something it used to know, Settings > Memory offers an export of your legacy memory, and that option disappears after 9 September 2026. Export it, then paste the missing part back into a chat and tell Claude to remember it. If any of it is health or identity detail, check the sensitive-topics setting first or it will not be saved. The deadline is close, so do it now rather than at the end of next week.

What memory cannot do for you

It cannot tell which of the things you mentioned actually mattered. That judgement is still yours, which is why the Topics list repays a read.
It cannot know when a fact has expired. A preference you held in March sits there looking current until you edit it.
Treat it as a convenience, not a filing system. Keep anything you would be upset to lose somewhere you control.

Claude changes most weeks. We read every release note and send you the parts that change how you work

One issue every morning, in plain English, for people who use AI at work.

Free forever. Unsubscribe anytime.

Common questions about Claude memory

Does my colleague see what Claude remembers about me?

No. Memory is per person. On Team and Enterprise plans an owner decides whether memory is available at all, but enabling it does not let anyone read your entries.

Is this the same as the context window?

No, and the two get confused constantly. The context window is how much text Claude can hold in one conversation, and it resets when that conversation ends. Memory is the small set of facts that survives. We cover the first in our guide to Claude context and long documents.

Does searching past chats need memory turned on?

No. They are separate switches in the same panel. Search reads back through conversations you still have. Memory is a short list of facts Claude wrote for itself. You can run either without the other.

Is memory available on the free plan?

Yes, and it is on by default. Searching past chats is the paid-plan feature. See what else the free tier includes in our guide to Anthropic’s free tier.

What should I actually do about all this?

Open Settings > Memory, read the Topics list, delete anything wrong or stale, and decide on the sensitive-topics toggle deliberately rather than by default. If you use projects, look inside those too, since each keeps its own memory. About ten minutes, once.

Sources

Anthropic — Use Claude’s chat search and memory to build on previous context
Anthropic — Claude release notes (memory across Cowork, editable topics and the sensitive-topics setting, 25 August 2026)
Anthropic — Use incognito chats

All settings and defaults above were checked against Anthropic’s help centre on 29 August 2026. Claude changes often; the settings panel is the authority.