ChatGPT, Claude or Gemini

Paste what ChatGPT, Claude or Gemini remembers about you, review it fact by fact, and your agent can answer from it. It is a one-time paste, never a sync.

Add it

Choose your assistant

In Knowledge, press Add a source (Add new on an Expert Twin’s page) and open Assistant memory. Pick the assistant.

The memory import form: the ChatGPT, Claude and Gemini tabs, the two ways to copy, the steps and an empty paste boxThe memory import form: the ChatGPT, Claude and Gemini tabs, the two ways to copy, the steps and an empty paste box
Knowledge › Add new
Copy what it remembers

Copy from settings is exact: open Settings → Personalization in ChatGPT, Settings → Capabilities → Memory in Claude, or Settings → Personal context → Saved info in Gemini, and copy what’s shown.

Ask it to write them out gives you a prompt (Copy the prompt) to paste into a new chat. That reply is the assistant describing itself, so it can include things it never stored. Read every line.

Paste it and review the facts

Paste into Paste what you copied. It splits into facts, one per bullet, numbered item or line, and each gets a tick. Untick anything wrong, private, or about someone else. You can leave a fact out but not rewrite it.

Three pasted facts, each with a tick, the line 3 of 3 facts will be added, the consent tick and the Add 3 facts buttonThree pasted facts, each with a tick, the line 3 of 3 facts will be added, the consent tick and the Add 3 facts button
Review before anything is stored
Agree and add

Tick I’ve read these and agree to hiy storing them, then press Add N facts. The tick covers what a memory can hold: health, sex life or sexual orientation, politics, religion, finances or relationships. See What you agree to.

Answers, not voice

An imported memory is a summary another assistant wrote in its own third-person register, so it feeds your agent’s answers like any other source but never how it sounds. If your agent’s tone drifts, this import isn’t why.

Warning

If your agent is public, anyone who chats with it can be shown one of these facts as an answer, quoted with the import’s title beside it as the source. Untick anything you wouldn’t say to a stranger.

In your sources the import shows a MEM badge, a title like “What ChatGPT remembers about you — 12 facts”, where it came from (“copied from ChatGPT settings” or “written by ChatGPT”) and “answers only, not voice”.

Remove it

Press Remove on its row in Knowledge. It leaves your agent at the next train, and until that train finishes your agent can still answer from it.

That removes hiy’s copy only. ChatGPT, Claude or Gemini keeps what it remembers until you clear it in its own settings.

Good to know

  • Only an Expert Twin offers it.
  • Nothing syncs. There’s no refresh button, so if an assistant learns something new, paste again.
  • One import holds up to 200 facts, 600 characters each.
  • The split and the review happen in your browser, so a fact you untick never reaches hiy’s servers.
  • New facts join your agent’s answers at the next train.
  • No assistant offers its memory over an API, so a paste is the only way in.

Was this page helpful?