ᛗMUNINN
GuidesPricing
Sign inTry Muninn
Back to home

Guide

How to transfer ChatGPT memory to Claude

Three paths, from a one-time paste to a shared workspace

ChatGPT and Claude don't share memory — each assistant has its own private store. To use the same context in both, you either move it once (quick) or set up a shared notes source both assistants can read (durable).

Step 1 — Get ChatGPT's memory out

OpenAI doesn't publish a memory export, but ChatGPT itself can list what it has. In any chat, ask:

"List everything you have stored in memory about me. Format it as a plain bulleted list I can copy."

Copy the response. Add anything from your Custom Instructions — those are separate from memory but do the same job.

Option A — Paste into Claude's Custom Instructions

The fastest way. In Claude, go to Settings → Profile and paste the list into the "What personal preferences should Claude consider?" field. Claude will honor it in every chat going forward.

Trade-off: it counts against Claude's context budget on every message, so keep it tight — a page or two, not everything.

Option B — Paste at the start of chats

If you only need the context in some conversations, start those chats with the list. Works, but you'll forget half the time and drift back to no-context Claude.

Option C — Shared notes workspace over MCP (recommended)

The durable answer is to stop putting facts in an assistant at all. Keep them in a notes workspace and let both assistants read it over the Model Context Protocol.

  1. Sign up for Muninn — a private notes workspace with a built-in MCP server.
  2. Paste your exported memory into a note called "About me".
  3. Connect Muninn to both Claude and ChatGPT — see the setup guide.
  4. In any chat, ask the assistant to search your Muninn notes for "about me" — or let it decide. Both assistants now read from the same source.

Update the note once; both assistants see the change. No copy- pasting between profiles, no hitting a memory cap.

Verifying the transfer worked

Ask Claude something that depends on a transferred fact:

  • "Based on what you know about me, would I like X or Y?"
  • "Give me a summary using my preferred tone and format."

If Claude leans on the transferred details, it's working. If not, check the Custom Instructions field (Option A) or that the MCP connector is enabled for the current chat (Option C).

Related reading

  • Does Claude have memory like ChatGPT?
  • ChatGPT saved memory is full — what to do
  • Connect Muninn to Claude and ChatGPT

Try Muninn

One memory for every assistant

Keep notes once. Let Claude, ChatGPT, and any MCP client retrieve and update them when you ask.

Get started