Claude's memory lets it remember context from past conversations so you don't have to reintroduce yourself every time. By default Claude starts each conversation from a blank slate, but with memory on, it carries context like your job, ongoing projects, and preferred tone across chats. This guide explains what memory is and how to turn it on, off, and manage it, based on the official docs. (As of June 2026 · Official: support.claude.com)
What Claude memory is
First, the premise. By default Claude starts fresh each conversation. Within a single chat it remembers everything said, but when the chat ends that context is gone and a new chat knows nothing of the last one. Memory solves this "start from scratch" problem by creating continuity across conversations.
Two features
Per the official docs, memory works in two ways.
- Search past chats — ask something like "find what we discussed before," and Claude searches your previous conversations to find and reference relevant details. This is available on paid plans (Pro, Max, Team, Enterprise) across web, Desktop, and Mobile.
- Auto-generated memory — Claude analyzes your chat history in the background and learns context like your job, projects, and tone, then reflects it in later conversations.
Turning it on and off
You can enable or disable both features yourself. Typically you'll find the "search and reference chats" and "generate memory from chat history" toggles under Settings → Capabilities or Memory. You can turn them off and back on anytime, and keep just one on if you prefer.
What it remembers · managing it
Auto memory extracts collaboration-helpful context like your job, ongoing projects, and how you like to work. Importantly, you stay in control: in the memory settings you can see exactly what Claude remembers and add, edit, or delete items. You can also say "remember this" or "forget that" explicitly mid-conversation, and ask "what do you remember about me?" to review the current contents.
Project memory · incognito
If you use Claude Projects, each project gets its own separate memory, so product planning, client work, and general tasks don't bleed together (see the Projects guide). And when you don't want anything saved, use an incognito chat — toggled via the ghost icon in the top-right. In that mode chats aren't saved and aren't added to memory. Incognito is available on all plans, including free.
Importing from other AIs
There's also an option to bring context you've built up in other tools like ChatGPT or Gemini. Claude generates an import prompt you use to fetch memory from the other account, then imports it into Claude (menu names and steps may vary by version).
Things to know
- Before you turned it on — memory doesn't cover chats from before it was enabled. Turn the feature off and back on and it starts fresh, without the previous memory.
- Deleting chats — deleting a conversation leads to the memory derived from it being cleaned up afterward.
- Not in API or Claude Code — memory is a claude.ai and Claude app feature. It doesn't apply to API calls or Claude Code.
- Recency — auto memory is a living synthesis rather than a fixed record; older details may fade over time.
Related
- Using Projects — persistent context via knowledge/instructions
- Keeping context in long chats
- Key features guide — Projects, Artifacts, Memory
Disclaimer: This article is based on Anthropic's official help docs (support.claude.com). Menu names, setting locations, behavior details, and edit limits may change by plan and version; some specific figures differ across sources and are intentionally not stated here. Check the official help center for the latest. This site is not affiliated with Anthropic.