Does Claude Have Memory? How to Turn It On and Use It

Yes. Claude keeps memory between conversations, it is available on the free, Pro and Max plans, and you switch it on in Settings under Memory by enabling the option to generate memory from chats. Unlike a chat window that starts blank every time, it carries your role, your working style and your current projects forward.

What makes Claude’s version distinctive is that you can read it. The memory is a set of individual entries you can open, edit and delete one by one, rather than a black box you have to interrogate by asking the model what it thinks it knows.

Turning it on

Memory lives in Settings under Memory, where the switch is labelled as generating memory from chats. The same screen holds the two controls that matter later.

• Pause stops Claude writing anything new while keeping everything it already has. This is the one to use when you are working on something you do not want folded into your profile.
• Reset permanently deletes every memory, including the memories held inside projects. It is not reversible, so treat it as a last resort rather than a tidy-up.

Availability has moved quickly. The current memory experience covers free, Pro and Max, and it has been rolling out to Team and Enterprise. If you are on a work plan and the setting is not where this describes, your organisation is most likely still on the older experience.

What it actually stores

Claude publishes the categories, and they are worth reading closely:

• your role, projects and professional context
• your communication preferences and working style
• your technical preferences and coding style
• your project details and ongoing work

Four categories, and all four describe the person typing. That is deliberate. Memory pays for itself by making the next answer better, and a fact about you applies to every conversation you will ever have. It is the reason Claude gets noticeably better at drafting in your voice after a few weeks, and also the reason it will not recall what a colleague told you in March.

Project memory is the part most people miss

Claude does not keep one undifferentiated pile. Each project gets its own memory space with its own summary, kept separate from your non-project chats, which share a common pool between them.

This is genuinely useful and slightly counterintuitive. Context you build inside a client project does not leak into an unrelated one, so Claude will not start applying one client’s house style to another’s deliverable. The flip side is that a preference you establish inside a project stays there. If you find yourself re-explaining the same working style in every new project, that is why, and the fix is to state it in a non-project chat so it lands in the shared pool instead.

Importing memory from ChatGPT or Gemini

Claude can take memory across from other assistants, which is the single fastest-rising question about AI memory right now. The flow is deliberately low-tech: you run a prompt in the assistant you are leaving, it produces a block of text summarising what it knows about you, and you paste that into Claude’s import field.

It is worth being clear about what moves. You get a summary of your preferences, your role, your ongoing projects and the instructions you kept repeating. You do not get the substance of the conversations that produced them, because that substance was never stored as substance. It was stored as a description of you.

That gap is not a flaw in the importer. It is an accurate reflection of what the source held.

What memory will not do for you

Two limits are worth knowing before you rely on it.

It is not indexed by person. You can ask Claude what it knows about you and get a coherent answer. Ask what it knows about a specific colleague and you are relying on whatever happened to be synthesized into a professional-context entry. There is no profile for Priya, because people are not one of the four categories.

It is scoped to Claude. Memory you build here does not reach ChatGPT, Gemini or anything else, and the import above is a one-time paste rather than a live sync. Every assistant keeps its own picture of you, and none of them hand it over to a rival.

Neither limit is a reason to leave the feature off. They are a reason to be deliberate about which things you expect it to hold. Preferences, context and working style belong in the assistant. Dated facts about the people in your working life need a store that indexes by person and does not depend on which assistant you are using this year. The full comparison across vendors is in what ChatGPT, Claude, Gemini and DeepSeek remember.

A worked example of the split

Say you have spent a quarter using Claude on a market-entry project. Its memory now knows you write in short paragraphs, that you prefer tables to bullet lists, that the project ships in October, and that you work in climate tech. Every one of those makes the next draft better, and none of them expires in a way that hurts you.

In the same quarter you had this conversation, which has no home in that store:

Call with Priya, 3 June. She has left the agency and is consulting solo, underpricing herself badly. Her mother’s scan came back clear. Wants an intro to someone who has run a two-sided marketplace. Said to check in after Lisbon in August.

Paste that into Claude today and you will get a warm, well-judged follow-up message straight away. Ask about Priya in November and there is nothing, because none of it was a fact about you. An app built for relationship memory files it under Priya, keeps the date, and answers from what you actually wrote rather than filling the gap with something plausible.

Key takeaway: Claude has memory on free, Pro and Max, enabled in Settings under Memory, and it stores your role, working style, technical preferences and project details as entries you can read and edit. Each project keeps a separate memory space while non-project chats share one pool, you can import a profile from ChatGPT or Gemini as a one-time paste, and nothing in it is indexed by person.

FAQ

How do I turn on Claude’s memory?

Go to Settings, open Memory, and enable generating memory from chats. The same screen lets you pause memory, which keeps what exists but stops new writes, and reset it, which permanently deletes everything including memories held inside projects.

Which Claude plans include memory?

The current memory experience is available on free, Pro and Max, and it has been rolling out to Team and Enterprise. If your work account does not show the settings described here, it is most likely still on the older experience.

Does Claude remember across separate chats?

Yes, with one important distinction. Chats that are not inside a project share a common memory pool, so context carries between them. Each project keeps its own separate memory space, so context built inside a project stays in that project.

Can Claude import my ChatGPT memory?

Yes. You run a prompt in ChatGPT to produce a summary of what it knows about you, then paste that into Claude’s import field. What transfers is a profile of your preferences and context, not the detail of your past conversations.

What is the difference between pausing and resetting Claude’s memory?

Pausing stops Claude writing new memories while keeping everything it already holds, so it is reversible. Resetting permanently deletes all memories, including project memories, and cannot be undone.

Can Claude remember details about my clients or colleagues?

Not reliably. Its memory categories cover your role, working style, technical preferences and project details, all of which describe you rather than the people you work with. For durable, person-indexed recall you need a store built around people, which is covered in AI that remembers conversations.

Intriq files dated notes under each person and answers only from what you wrote. Start with the personal CRM hub or read can AI remember people for you.