Claude Memory: How It Works and How to Turn It Off

One of the things that used to frustrate business owners about AI assistants was starting from zero every time. You'd explain your business, your clients and your tone of voice, get a great answer, and then the next conversation had forgotten all of it. Claude's memory is Anthropic's answer to that, and it's worth understanding before you decide whether to leave it on.

Here's how Claude memory works according to Anthropic's help pages, how to switch it on, off or pause it, how it differs from searching past chats, and what I'd never let it store.

What is Claude memory?

Claude can build up memory from your chats and carry that context into new conversations. Anthropic describes it as turning Claude from a chat window that forgets everything into a collaborator that builds understanding over time.

Claude saves memory as a set of individual topics as you chat, rather than summarising conversations after they end. So if you mention that a deadline has moved, your next conversation already knows. It saves things on its own, and you can also tell it to remember something directly.

Is Claude memory on by default?

It depends on your plan. Anthropic says memory is on by default for the Free, Pro and Max plans on the web, Claude Desktop and the Claude mobile apps. On Team and Enterprise plans, owners decide whether memory is available, and it stays off for each member until they turn it on. It's one of the admin controls covered in my guide to what the Claude Team plan includes.

How do you turn Claude memory on or off?

Go to Settings, then Memory, and switch on Generate memory from chats. When you switch it off, you'll get two choices:

• Pause memory: Claude keeps what it already remembers but won't use it or make new memories. Chats while it's paused won't be added later if you switch it back on.
• Reset memory: permanently deletes all memories, including project memories. This can't be undone, so Claude starts from scratch if you turn it on again.

If you're unsure, pause rather than reset. You can always reset later; you can't get a reset memory back.

Can you turn off memory for one chat?

Yes, and this is the setting I'd use most. Leave memory on for your account, and switch it off for a single chat when the subject is sensitive or simply irrelevant to your normal work. Start a new chat, click + in the message box and turn off Memory. Claude won't use what it remembers in that chat, and nothing from that chat gets saved to memory.

You have to do this before you send the first message; once the chat has started, the setting is fixed. Unlike an incognito chat, this chat still appears in your history.

What's the difference between memory and searching past chats?

They're related but separate. Searching past chats lets you ask Claude to find something from an earlier conversation, such as asking what you discussed about a topic, and it pulls in the relevant context. Anthropic says that's available on the Pro, Max, Team and Enterprise plans, it's on by default once it reaches your account, and you'll see it as a tool call in the chat. You can switch it off with the Search and reference chats toggle in Settings, Memory.

Memory goes a step further: Claude builds up what it knows about you and your work without you asking it to look anything up.

What is an incognito chat in Claude?

Incognito chats are available on every plan. When you start a new chat outside a project, click the ghost icon in the top right corner. The conversation isn't saved to your chat history, and Claude won't pull from it when searching past chats. Anthropic notes that on Team and Enterprise plans, incognito chats are still included in data exports and follow your organisation's retention policies.

Do Claude Projects have separate memory?

Yes. Each project has its own memory space and project summary, so what Claude learns in one project stays focused there and separate from other projects and your everyday chats. For a business, that's useful: work for one client doesn't bleed into another. You can manage memory partly by moving chats into the right project.

Does memory work in Claude Cowork?

When Cowork runs in the cloud, it shares memory with your chats, so a task already knows who your clients are and how you like things done, and what comes up in the task carries back. Cowork sessions that run locally on your computer don't use memory.

How do you see and edit what Claude remembers?

Go to Settings, then Memory. Everything Claude remembers is listed under Topics. Select any topic to read it, use the edit icon to change it, or select Delete to remove it. Anthropic says a fix in one topic applies to every conversation from then on, so correcting an out-of-date detail once is enough.

Anthropic also says some information is never saved to memory, even if you ask, including government ID numbers, criminal history, financial account numbers and immigration status. Certain sensitive topics are only saved if you choose to switch on Include sensitive topics in memory, and on Team and Enterprise plans owners can't view or edit a member's individual memories. Anthropic also offers an experimental way to import memory from other AI assistants or export your Claude memory as a backup.

What should you never let Claude remember?

Memory is only as safe as what goes into it. My rules for business owners:

• No passwords or logins, ever, in any chat.
• No client personal details such as addresses, phone numbers or health information. Use a chat with memory off, or incognito, if you must work with them.
• Nothing under a confidentiality agreement unless your account and settings have been approved for it.
• No guesses about people. If you vent about a client, Claude may remember it.

And it's worth a regular look at what's been saved. Here's a prompt to help.

Copy this prompt

Tell me what you currently remember about me and my business, grouped by topic.
Flag anything that looks out of date, sensitive or about a specific person, so I can decide whether to keep it.

Then use this to set memory up well from the start:

Copy and paste this prompt

Please remember these things about my business for future chats:
1. I run [business name], a [type of business] serving [who you serve].
2. I write in [UK / US] English, in a [tone] tone, and never use [words].
3. My main jobs for you are [list them].
Don't remember client names or personal details from our chats unless I ask you to.

Tools that work alongside Claude memory

Memory works best alongside two other Claude features: Instructions for Claude, which apply to every conversation, and project instructions for each client or job. If you'd like a bigger head start on your tone, an AI voice bank built from your old content gives Claude real examples to learn from. And if you're still choosing between assistants, my comparison of the right AI tool for each job covers how others handle it.

If you'd like a sensible memory and privacy set-up agreed for your whole team, this is how I work with business owners.

Frequently asked questions

Does Claude have memory?

Yes. Claude can build memory from your chats and use it in new conversations. It saves memory as individual topics as you chat, and you can also ask it to remember something.

How do I turn off Claude memory?

Go to Settings, then Memory, and switch off Generate memory from chats. Choose Pause memory to keep what's saved, or Reset memory to delete everything permanently.

Is Claude memory on by default?

Anthropic says memory is on by default for Free, Pro and Max plans. On Team and Enterprise plans, owners control it and it's off for each member until they turn it on.

Can I use Claude without memory for one chat?

Yes. Start a new chat, click + in the message box and turn off Memory before your first message, or use an incognito chat, which isn't saved to your history.

Do Claude Projects share memory?

No. Each project has its own memory space and project summary, separate from other projects and your everyday chats.