Move your ChatGPT memory to Claude or Gemini

Import your ChatGPT memory into Claude

Claude has a built-in Import Memory feature on the web and desktop apps. It works in three steps:

1. In Claude, open Settings and find Import Memory. Copy the prompt Claude shows you.
2. Paste that prompt into ChatGPT. It answers with what it remembers about you in one block.
3. Paste ChatGPT's answer back into Claude and click Add to memory. Review the new entries in Claude's memory panel.

Anthropic calls the feature experimental. Claude may not keep personal details unrelated to work, and only memory comes over, not chat history. The details are in Anthropic’s import guide.

The import is a one-time copy into Claude. ChatGPT keeps its own memory, and the two drift apart from there. If you use both, or add Gemini later, keep the context that matters in notes every assistant can read.

Review the context you want to carry over

ChatGPT memory, conversation history, and project files are different kinds of context. Review the controls available on your account and inspect anything you copy or export. A memory summary can help you collect useful facts, but it may omit details or include mistakes. Check the facts before giving them to another assistant.

OpenAI documents memory settings that vary by account and surface. Check its personalization documentation before following a transfer guide. Saving selected facts as notes is a deliberate transfer of context; it does not copy every conversation or synchronize the assistants’ native memory.

Keep maintained context in shared notes

Write the projects, preferences, decisions, and context you want to maintain into notes you can review. Hjarni stores Markdown notes and exposes them through its built-in MCP server. Compatible connected assistants can search, read, create and update those notes through permitted tools.

When you add another compatible assistant, connect it to the same saved notes and ask it to retrieve the relevant context. The notes persist in Hjarni across conversations. Connecting does not automatically load them into every chat or move either product’s built-in memory.

Import your ChatGPT memory into Hjarni instead

Once ChatGPT is connected to Hjarni, it can write the notes itself. Paste this into ChatGPT:

Hjarni, save what you remember about me as notes. One note per topic: my work, my projects, my preferences, the people I mention. Show me the list first and wait for my OK before saving.

Correct the list before you say yes. ChatGPT's summary can miss things or get them wrong. Then open the notes in Hjarni to check them. The same prompt works in Claude, so you can pull in what Claude remembers too.

Now ask Claude, ChatGPT or Gemini CLI about one of those topics. Each one reads the same notes, and when you fix a note, every assistant sees the fix. See the long-term memory with MCP guide for how retrieval works.

Keep the context you want to reuse in notes you can inspect, correct, and ask your connected assistants to retrieve.