Coredence

Guide

Switching AI assistants without starting over

The hardest part of moving to another assistant is not the features. It is the months of context the old one built up about you — and the feeling of starting from a stranger.

Ahmet Karaca ·

What actually moves

  • Your chats, as files. ChatGPT (Settings › Data controls › Export data) and Claude (Settings › Privacy › Export data) both send an email link to a download of your conversations. The link expires after 24 hours.
  • Your memories, as text. Neither ChatGPT nor Claude exports its memory as a file. The newer assistants solve this from the receiving side: they give you a prompt to paste into the old one, which lists what it remembers, and you paste the answer back.
  • Your instructions, by hand. Custom instructions in ChatGPT, Instructions for Gemini, project instructions in Claude — copy them across yourself.

Moving to Claude

Claude has a memory import under Settings › Memory › Start import (also at claude.com/import-memory). It gives you a prompt for your old assistant; paste the answer and choose Add to memory. It keeps working style, preferences, projects and tools, favours work over unrelated personal detail, and merges with what Claude already knows.

Anthropic marks it experimental and says it may not always take everything in. Read Claude's memory afterwards (Settings › Memory) and correct what came across wrong.

Moving to Gemini

Gemini has Settings & help › Import memory to Gemini, which works the same way for memories, and it can also take a chat-history export as a .zip file. Imported chats are saved to your activity. The feature needs a personal account and an age of 18 or over, and it is not offered in the EEA, Switzerland or the UK.

What gets lost on the way

An import carries over what the old assistant chose to write down, in its words. Whatever it got wrong comes along too, and nothing says which memories rested on one remark and which on a hundred.

And the next move starts the problem again. Each assistant keeps its own picture of you, and none of them hands it to the others.

  • Keep your chat exports. They are the only complete record, and the only thing you can read again later.
  • Read the memory list before you paste it anywhere. It is a good moment to delete what is no longer true.
  • Write down the few things you correct most often. Those are the ones you will otherwise explain again.

Where Coredence fits

Coredence keeps that picture on your side instead of inside an assistant. It reads your ChatGPT and Claude exports, notes and a journal if you want, and consolidates them into memories about how you think and work — each one tied to the traces it grew from, so you can see what it rests on and correct it.

Because it is one file on your computer, switching assistants does not reset it. Coding assistants read it directly. For Claude, ChatGPT and Gemini, Settings › Assistants › “Copy and open” copies your strongest memories at the length each one accepts and opens the right place; the imports above become a single paste, with no old assistant in between. Nothing is sent from the app, and you can repeat it whenever your mind grows.

You can try the real app on a made-up person's mind, in your browser, without installing anything.

Try it in your browser