Blog

How to import ChatGPT memory into Claude and Gemini

Connect Configure to both assistants, and what ChatGPT knows about you is there in Claude or Gemini. It works in every other direction too.

Configure · · 5 min read

In short

  • To move ChatGPT memory to Claude, add Configure Memory in ChatGPT, ask ChatGPT to save what it knows about you, then add the Configure connector in Claude.
  • The same steps work in every direction and for other assistants too, including Gemini, Grok, Perplexity and Muse.
  • Your memory goes into a profile you own. Every assistant you connect reads from it, so they don’t drift apart the way they do after a one-time copy.
  • Claude and Gemini have built-in imports too, but they copy once. ChatGPT has no import at all.

To import your ChatGPT memory into Claude, add the Configure Memory app in ChatGPT and ask ChatGPT to save what it knows about you. Then add the Configure connector in Claude and sign in with the same phone number. It takes a few minutes, with nothing to copy or paste. Your memory goes into a profile you own, which keeps both assistants in step. The same steps work for Gemini.

Where do you add Configure?

Inside each assistant, in its own connector settings. We build Configure, which is infrastructure for connecting agents, apps, and chats, and carrying your memory between them is one of the things it does. Start with the two assistants you are moving between.

Why is your memory worth moving?

Because it is most of what makes an assistant useful. After months of use, ChatGPT knows your job, your projects, the people you mention and how you like things written. Move to an assistant with a better model and none of that, and you still get the answer anyone would get. You also spend the first weeks explaining yourself again.

That memory is yours. You wrote it one conversation at a time, and you should be able to take it wherever you work next.

How do you import ChatGPT memory into Claude?

  1. Add Configure Memory in ChatGPT. Open the Configure Memory app in ChatGPT and connect it.
  2. Sign in. ChatGPT sends you to Configure. Enter your phone number and the code you are sent, then approve access. You land back in ChatGPT.
  3. Ask ChatGPT to save what it knows. Start a new chat and send the request below. ChatGPT saves each thing it knows as its own fact, up to 100 at a time.
  4. Add the Configure connector in Claude. In Claude, open Customize, then Connectors, and add Configure.
  5. Sign in with the same number. Choose Sign in now, enter the same phone number and the new code, and approve access.
  6. Check it. Ask Claude, “What do you know about me from my Configure profile?”
Save what you know about me to Configure: my saved memories, and what our past chats say about my work, projects, goals, hobbies, preferences and the tools I use. One fact per item.

From here Claude can look up what it needs about you, and save a new fact when you ask it to keep one. ChatGPT can do the same, so the two no longer need a second transfer. Configure Memory for Claude lists what the connector can do and what it receives.

How do you move memory in the other directions?

The same way. Connect Configure to both assistants, then send that request to the one that already knows you. Each row opens the steps for that assistant, with pictures of its screens. They are also on one page: How to add Configure to your assistant.

AssistantHow Configure is addedSteps
ChatGPTThe Configure Memory app
ClaudeThe Configure connector, listed in Claude’s directory
GeminiA custom app, added from a computer. Personal accounts in the US, in English
GrokA custom connector
PerplexityA custom connector. The free plan does not show the option
MuseMuse has no box for an address, so you ask it to connect
Any other assistantA custom connector, sometimes called a remote MCP server

Muse works the way any assistant without connector settings can. If it can make web requests, give it https://connect.configure.dev and it sends back a link for you to approve. Connect over MCP has the details.

You are not limited to two. Connect a third assistant and it reads the same profile, and anything you ask it to save is there for the first two.

What is different about doing it this way?

With ConfigureA built-in import
After the moveBoth assistants read and add to the same memoryA one-time copy. The two drift apart from the first day
DirectionsAny. Every connected assistant can save and readInto Claude or Gemini only. ChatGPT has no import
Copying and pastingNone. The assistant saves straight to your profileRun a prompt, copy the answer, paste it into the other assistant
A third assistant laterConnect it. It already knows youStart again with a new prompt and a new paste
Seeing what was keptOne page with every memory and which assistant saved itInside each assistant’s own settings
Apps beyond assistantsWorks in any product that supports ConfigureNo

A copy made in March is out of date by May. And most people who try a new assistant do not leave the old one, so they end up using both. Keep the memory in one place both can reach and there is nothing to drift.

What about Claude’s built-in import?

Claude has one, and it is fine for a single move. In Claude’s Settings, select Start import, run the prompt it gives you in ChatGPT, paste the answer back, and choose Add to memory. Anthropic’s help article has the steps. Anthropic marks it experimental and says an import may not always be taken in. It also notes that Claude may not keep personal details unrelated to work.

Gemini has a similar import, which is not offered in the European Economic Area, Switzerland or the UK. ChatGPT has no official way to import memory from another assistant, so a connector is the only way to move memory into ChatGPT.

What should you check afterwards?

Start by reading what came across. Your Configure profile lists every memory and which assistant saved it, and you can edit or delete anything that is wrong.

Then look for gaps. ChatGPT decides what to save, and it sometimes leaves things out. If something important is missing, ask ChatGPT to save that one thing, or tell the new assistant and ask it to keep it.

Some things are left out on purpose. Configure does not store a fact it recognizes as a password, a card number, a government ID number or a health detail.

Old or wrong memories in ChatGPT come along too. If you clear them out first, under Settings, then Personalization, then Memory, you get a cleaner start.

Frequently asked questions

Does this move my chats?

No. It moves what the assistant knows about you, meaning your preferences, your projects and the details it has kept. The conversations stay where they are.

Can I switch from ChatGPT to Claude and keep using both?

Yes, and that is the point. Once both are connected, a fact you ask one to save is there for the other.

Do I have to paste anything?

No. The assistant saves to your profile through the connector. You send one request and it does the rest.

Can I undo it?

Yes. From your profile you can delete any memory, disconnect an assistant, or delete the whole profile.

Does it work with apps as well as assistants?

Yes. Products that support Configure show a connect button, and they start from what you have already shared. If you build one, the quickstart shows how. For how the assistants’ own memories compare, see Which AI has the best memory?

Add a connect button to your product

The quickstart takes one page, end to end: a chip in your app, Configure's hosted page, and your agent reading the user's context on the first turn.

Read the quickstart

About Configure. Configure is infrastructure for connecting agents, apps, and chats. People bring the context they already have, and the products they approve start from it. More about Configure.

More from the blog

All posts