Claude Memory (2026): How It Works, Import Memory & Should You Turn It On?
How Claude's memory saves topics from your chats, how to import memory from ChatGPT, and when to switch it off.

Claude memory is the feature that lets Claude carry what it learns about you from one chat into the next. The 2026 version saves that knowledge as individual topics while you talk, lets you read and edit every topic in Settings, and is on by default for Free, Pro and Max.
Below: how it works now, how to import memory from ChatGPT, how to turn it off, and where it stops being enough.
What is Claude memory?
Claude memory is a store of facts about you and your work that Claude builds from your conversations and reads at the start of new ones. Mention once that you run growth at a fintech and write in British English, and the next chat already knows.
Guides that say Claude forgets everything between chats are out of date. Anthropic shipped memory to Team plans on September 11, 2025, to Max on October 23, 2025 (with Pro following over the next two weeks), and to every plan including free on March 2, 2026, according to its release notes.
What's new in Claude's memory feature?
The 2026 version stores memory as separate, editable topics instead of one summary rewritten every day. Two updates got it there:
- July 10, 2026: memory became "a set of individual, categorized entries" that Claude reads and updates, replacing the daily summary.
- August 25, 2026: memory started working across chat and Cowork in the cloud, every item became editable under Topics in Claude's memory settings, and a new setting keeps sensitive topics out unless you opt in.
A small number of Team and Enterprise organizations still run the legacy version, which lives under Settings > Capabilities and refreshes a summary every 24 hours (Anthropic, 2026). If you see Memory under Capabilities, that's the one you have.
How does Claude memory work?
Claude decides during a conversation what's worth keeping and saves it as a topic right away, rather than waiting for the chat to end. You can also say "remember this" to save something directly, or "forget that" to drop it.
What you control, per Anthropic's memory guide:
- Topics: in Settings > Your Claude, select Manage next to "Your memory" to see everything Claude remembers (in the older settings layout, this page is Settings > Memory). Open a topic to read it, edit it, or delete it.
- Sensitive topics: health, politics and similar subjects stay out of memory unless you turn on "Include sensitive topics in memory". Government ID numbers, financial account numbers, criminal history and immigration status are never saved, even if you ask.
- Chat search: on paid plans, Claude can also search and reference your past chats. That has its own toggle, "Search and reference chats", in the same panel.
Since the 2026 update, memory covers everyday context as well as work; only the legacy version focuses on work alone.
How does memory work in Claude Projects?
Each Project keeps its own memory. Inside a Project, Claude uses that Project's memory plus your general memory, but never memory from other Projects. To wall a Project off completely, for example client work you don't want bleeding into anything else, turn off "Use account memory" on the Project's page.
Can Claude import memory from ChatGPT?
Yes, as a one-time import rather than a live sync. Anthropic's import page gives you a prompt to paste into ChatGPT, Gemini or any other assistant; it answers with everything it has stored about you, and you paste that back into Claude's memory settings. Claude splits it into individual memory entries you can review.
Import works on Free, Pro, Max, Team and Enterprise, on the web and in Claude Desktop, and Anthropic still labels it experimental: Claude "may not always incorporate imported memories successfully" (Anthropic, 2026). Whatever you tell ChatGPT after the import stays in ChatGPT. For the full switch, see how to move your context from ChatGPT to Claude.
Should you turn on Claude memory?
Yes, if Claude is the AI tool you use most. It removes the routine re-introductions, you can see and correct every topic it keeps, and you can opt out of any single chat.
Leave it off, or use the per-chat switch, when:
- you want a clean answer that isn't shaped by your history, such as a second opinion or a test of how Claude answers a stranger;
- you handle client or confidential work and don't want it surfacing in unrelated chats (Project-only memory solves most of this);
- you're on a Team or Enterprise plan where your admin hasn't enabled it. Memory isn't available to organizations with HIPAA, public-sector or custom data retention agreements.
One habit makes memory far more useful: open Settings > Your Claude once a month and delete what's no longer true. Memory is inferred from your chats, so editing it means correcting Claude's guesses, not curating a profile you wrote.
What doesn't Claude memory do?
It doesn't leave Claude. Your topics live in your Claude account, and ChatGPT, Cursor and Gemini each build their own separate picture of you from scratch. That's a different kind of gap from the one memory solves, and it's the one personal context is about.
It's also reactive. Memory knows what you happened to mention, so it starts empty and fills slowly, and it doesn't know about the email thread or Notion doc you never pasted in. Connecting Gmail or Notion as a connector gives Claude a place to search, not a summary of who you are: it finds what matches the query and misses the planning doc that used different words.
Claude memory vs a portable context layer
Claude memory handles continuity inside Claude. A context layer like Unabyss handles the foundation underneath every tool: it builds your personal context from Gmail, Google Calendar, Notion, Slack and GitHub, keeps it current, and serves the same version to Claude, ChatGPT and Cursor over MCP.
The two stack. Keep Claude memory on for what you discuss in Claude, and add a layer when you notice you're explaining the same project to three different assistants. In an August 2026 Unabyss benchmark, giving an assistant structured context on top of its connected tools cut cost by 36% and time by 38%, and the quality of accepted output rose from 7.7 to 8.9 out of 10 (What Structured AI Context Is Worth). For a shortlist of options, see the best AI memory tools, or read the Unabyss vs LLM memory comparison.
FAQ
How do I import memory into Claude?
Go to Settings > Your Claude and select Start import. Copy the prompt Claude shows you, paste it into ChatGPT or your other assistant, then paste its answer into the text box and click Add to memory. New entries usually appear shortly after; See what Claude learned about you opens a chat to check them.
How do I turn Claude memory on or off?
For your whole account, use the "Memory from chats" toggle in Settings > Your Claude (called "Generate memory from chats" in the older layout). Turning it off gives you two choices: pause, which keeps existing memory but stops using and adding to it, or reset, which permanently deletes all memories, including Project memories. For one chat, click + before your first message and switch off Memory; you can't change it once the chat starts.
How do I get Claude to respond with no previous context?
Start an incognito chat with the ghost icon in the top right corner. Incognito chats aren't saved to memory or chat history, and Claude doesn't search them later. More options are in how to give Claude persistent context.
Can I export my Claude memory?
Yes. Ask Claude to "write out your memories of me verbatim, exactly as they appear in your memory" and save the result to a file, as Anthropic's import and export guide suggests. You can also read every topic under Settings > Your Claude > Manage.
Want every AI tool to know what Claude knows? See how Unabyss works.