Claude memory: what Claude remembers from one chat to the next

Updated 30 October 2026. Every feature described here comes from Anthropic's official help centre, read on 30 October 2026. Anthropic changes these settings: check the labels on your account.

Yes, Claude can remember from one conversation to the next, through two separate features. Memory (“Memory from chats”) keeps topics about you and your work; it is on by default on the Free, Pro and Max plans. Chat search (“Search and reference chats”) lets it find a past chat; it is limited to paid plans. Both are controlled in Settings → Your Claude, and both only serve Claude.

What Claude's memory keeps

• Topics, not chat summaries. Claude saves memory as a set of topics as you chat: your role, your projects, your professional context, your communication and technical preferences.
• One memory per project. Each project has its own separate memory; your account-wide memory stays available inside projects.
• What you can see and fix. In Settings → Your Claude, the Manage button next to “Your memory” shows the topics kept; you can edit or delete them, and the change applies to future conversations.

Source: Anthropic's article on chat search and memory.

Depending on your plan

• Free, Pro, Max: memory is on by default.
• Chat search: paid plans only (Pro, Max, Team, Enterprise), on web, desktop and mobile.
• Team and Enterprise: organisation owners make memory available; it then stays off for each member until they turn it on.

Turn it on, off, or keep a chat out of memory

1. Open Settings → Your Claude. In the older settings layout, it's Settings → Memory.
2. Memory: the “Memory from chats” toggle (older layout: “Generate memory from chats”).
3. Search across past chats: the “Search and reference chats” toggle.
4. For a single conversation: an incognito chat is neither remembered nor searched, on every plan. You can also turn memory off for one chat from the “+” menu, before sending the first message.

For guidance that applies to all your conversations, use instructions instead: Settings → Your Claude → Instructions. You write those, whereas Claude maintains its memory.

Import and export

Claude can take in a memory from another AI: Settings → Your Claude → Start import, paste the text, then Add to memory. Per Anthropic, import is available on Free, Pro, Max, Team and Enterprise, on the web and the desktop app; it is experimental, and Claude prioritises work-related context. To export, open Manage next to “Your memory”, or ask Claude to write out verbatim what it remembers about you.

The step-by-step from ChatGPT is in transfer your ChatGPT memory to Claude.

What Claude's memory doesn't do

• It doesn't leave Claude. ChatGPT, Cursor, Perplexity or Gemini CLI can't reach it: each AI has its own memory, tied to its product.
• You don't pick every topic it keeps. You can fix or delete afterwards, but Claude decides what is worth keeping.
• Claude Code has its own. The command-line tool relies on CLAUDE.md files and an auto memory local to your machine.

If you only use Claude, all of this is enough, and it is well designed. The need for something else shows up when you move between AIs. For Claude Code, see its memory and its limits.

Share the same memory with your other AIs

Claude accepts custom connectors: remote MCP (Model Context Protocol) servers, added in Customize → Connectors, including on the free plan (one connector). A memory like Pryzm Memory plugs in there: Claude saves to it and searches it, and ChatGPT, Cursor or Claude Code connected to the same account find the same memories.

In that case, turn off “Memory from chats” if you want a single source of truth. The steps are in our custom connector guide, and the proof that it works in the 2-minute test.