Claude remembers you now. Sometime in the last year, Anthropic quietly turned Claude from a chatbot that forgot everything the moment you closed the tab into one that keeps a running set of notes about who you are, what you're working on, and how you like things done. For a lot of people that's genuinely useful - you stop repeating yourself. For others it's unsettling: what exactly is it storing, and how do you make it stop? This is the plain-English, click-by-click guide to controlling Claude's memory - how to see what it knows, edit it, switch it off, wipe it completely, and chat with zero trace when you need to.
Quick answer: Claude's memory is ON by default on Free, Pro, and Max. To control it, go to Settings > Memory. There you can Pause it, turn off the two toggles ('Generate memory from chats' and 'Search and reference chats'), edit or delete individual memories, or hit Reset to permanently delete everything. One thing that surprises people: deleting a chat does NOT delete the memory it created - memories are stored as separate entries, so you have to clear them in Settings > Memory. For a truly private conversation, use an Incognito chat (available on all plans).
Yes, Claude Has Memory Now - and It's On by Default
First, the part that catches people off guard: Claude's memory isn't something you have to switch on. As of 2026 it is the default experience for new users, and existing Free, Pro, and Max users were migrated into it. Two settings ship turned ON out of the box: 'Generate memory from chats' (Claude writes notes about you based on your conversations) and 'Search and reference chats' (Claude can look back at your past conversations). Anthropic's own marketing once described memory as 'fully optional,' but the accurate way to describe the current state is opt-out: it's already working unless you go and turn it off. That's not sinister - it's how most assistant memory features work now - but it does mean the responsibility to review it is on you.
What Claude Actually Remembers About You
In mid-2026 Anthropic changed how memory works under the hood. It used to keep a single rolling daily 'summary' of you. Since July 2026, it instead keeps individual, categorized entries - discrete notes like 'prefers concise answers,' 'working on a Next.js app,' or 'based in India' - that Claude reads and updates as you chat. That's a meaningful shift for two reasons. First, it's more transparent: you can open the list and see specific facts rather than one vague blob. Second, it's more persistent: because each memory is its own entry, it survives even after you delete the original conversation that created it. Memory is also scoped - each Project you create has its own separate memory space, so work context doesn't bleed into your personal chats. On mobile, Claude can additionally hold onto things like health and fitness details if you share them. The takeaway: memory is now a real, structured record, not a fuzzy impression.
How to See Everything Claude Has Stored
Before you delete anything, it's worth actually looking at what's there - it's usually less dramatic than you fear, and occasionally more revealing than you'd like. Here's how to review and tidy it:
- Open Settings > Memory. This is the control center for everything memory-related, on web and mobile.
- Read the list of entries. Each stored memory appears as its own line - a specific note Claude has written about you or your work.
- Edit an entry: select it and use the 'Tell Claude what to change or remove' box to correct or refine what it says. Useful when a memory is outdated or slightly wrong.
- Delete a single entry: select it and choose Delete to remove just that one note, leaving the rest intact.
- Check your Projects: each Project has its own memory, so review those separately if you use them.
How to Turn Claude's Memory Off
If you'd rather Claude stopped building a profile of you, you don't have to nuke what's already there - you can simply stop it going forward. In Settings > Memory you have a few levels of control:
- Pause - the gentlest option. Claude keeps its existing memory but won't use it or create new memories until you unpause. Good for a temporary break.
- Turn off 'Generate memory from chats' - Claude stops writing new notes about you from your conversations.
- Turn off 'Search and reference chats' - Claude stops looking back at your past conversations for context.
- Turn both off for the closest thing to the old, forgetful Claude - it neither records new memories nor references old chats.
- On Enterprise, a separate 'legacy memory' experience may live under Settings > Capabilities instead; admins may also control availability org-wide.
How to Delete All of Claude's Memory - Permanently
If pausing isn't enough and you want a clean slate, Claude has a full reset. Go to Settings > Memory and choose Reset. This permanently deletes all of your memories, including memories stored inside Projects, and you'll be asked to confirm with a 'Reset memory' step. Take the warning seriously: it cannot be undone, and there's no trash can to recover from. The single most important thing to understand here - and the reason many people think they've cleared their data when they haven't - is that deleting a conversation does not delete the memories that conversation created. Since memories are now standalone entries, wiping a chat leaves its notes behind. If your goal is to remove what Claude knows, you must delete the entries in Settings > Memory (individually) or Reset them (all at once); clearing your chat history alone won't do it.
Incognito Chat: Talk to Claude Without Saving Anything
Sometimes you don't want to manage memory at all - you just want one conversation that leaves no trace. That's what Incognito chat is for, and it's available on every plan, including Free. Start a new (non-Project) chat and look for the ghost icon near the top-right to switch it to Incognito. Those conversations aren't saved to your chat history and aren't used to create or update memory. It's the right tool for anything sensitive - a health question, something about your finances, a work matter under NDA - where the cleanest approach is simply not to record it in the first place.
Does Claude Use Your Memory or Chats to Train Its Models?
This is the question people really mean when they ask about deleting memory, so let's separate two different things. Memory is a personalization feature - a set of notes Claude keeps so it can be more useful to you. Whether your conversations can be used to improve Anthropic's underlying models is a separate matter, governed by your Privacy settings, and the defaults for consumer (Free/Pro/Max) accounts have shifted over time. So the honest guidance is: turning memory off stops personalization, but it's the Privacy/data controls - not the memory toggle - that determine training. If this matters to you, open Settings > Privacy, review the data controls there, and choose what you're comfortable with. And for anything genuinely sensitive, use an Incognito chat rather than relying on settings alone. Two switches, two different jobs: Memory controls what Claude remembers about you; Privacy controls what Anthropic may learn from you.
| Action | What it does | Reversible? |
|---|---|---|
| Pause memory | Keeps memories but stops using/creating them | Yes - just unpause |
| Turn off the two toggles | Stops new memories and stops referencing old chats | Yes - toggle back on |
| Delete an entry | Removes one specific stored note | No - that note is gone |
| Reset memory | Permanently deletes ALL memories, including Projects | No - cannot be undone |
| Incognito chat | This chat isn't saved or turned into memory | N/A - nothing was stored |
- Claude's memory is ON by default on Free, Pro, and Max - it's opt-out, not opt-in.
- Since July 2026 it stores individual, categorized entries you can view and edit in Settings > Memory.
- Deleting a chat does NOT delete its memory - clear entries in Settings > Memory instead.
- Reset permanently wipes everything (including Project memory) and cannot be undone.
- Incognito chat (all plans) is the simplest way to talk to Claude with nothing saved.
- Memory and model training are controlled separately - check Settings > Privacy for training/data controls.
01How do I delete Claude's memory?
Go to Settings > Memory. To remove one item, select it and choose Delete. To erase everything, choose Reset and confirm with 'Reset memory' - this permanently deletes all memories, including those inside Projects, and cannot be undone. Note that deleting your chat history does not delete memories; you have to clear them here.
02Is Claude's memory on by default?
Yes. As of 2026, memory is the default experience on Free, Pro, and Max, and existing users were migrated into it. Both 'Generate memory from chats' and 'Search and reference chats' ship turned on, so it's opt-out - you have to go to Settings > Memory to turn it off.
03If I delete a conversation, does that delete what Claude remembers?
No - and this trips up a lot of people. Since mid-2026 Claude stores memories as separate, standalone entries, so deleting the original chat leaves its notes behind. To actually remove what Claude knows, delete the entries in Settings > Memory or use Reset to clear all of them.
04How do I turn Claude's memory off without deleting it?
In Settings > Memory, use Pause (Claude keeps existing memories but stops using or creating them), or turn off the two toggles - 'Generate memory from chats' and 'Search and reference chats' - to stop new memories and stop it referencing past conversations, while leaving current memories in place.
05Can I chat with Claude without it saving anything?
Yes. Use Incognito chat, available on all plans including Free. Start a new non-Project chat and switch it to Incognito using the ghost icon near the top-right. Those conversations aren't saved to history and aren't used to build memory - ideal for sensitive topics.
Claude's memory is a genuinely useful feature once you know how to steer it - the point isn't to fear it, it's to run it on your terms. Review what it stores, correct what's wrong, pause or wipe it when you want a clean slate, and drop into Incognito for anything private. If you want that kind of control across more than one AI, LumiChats lets you use many leading models - Claude, GPT, Gemini and others - under one login at a pay-per-day price, so you can pick the right assistant for each task and keep your data habits deliberate rather than default.
