Getting your conversations sooner

Requesting an export from ChatGPT is the right way to get your data, and it is slow. OpenAI’s own confirmation email says:

We have started preparing your data export, this process may take a few days. We will email you when it is ready to download

This page covers what to expect, the trap that costs people their export, and an optional faster route.

First: request it now

The single most effective thing is to start the clock before you need to. Settings → Data controls → Export data, at chatgpt.com/#settings/DataControls.

Two things about that process are not obvious and both cost people their export:

The download link expires 24 hours after the email arrives. This is where the widely repeated “24 hours” figure comes from. It is a deadline, not a wait. Miss it and you start again.

Only your most recent request is fulfilled. Requesting a second time because the first felt slow silently cancels the first. Ask once, then wait.

When it arrives, save the .zip into your inbox folder — data/inbox/ by default — and Mind Archive imports it on its own.

The faster route

scripts/browser/chatgpt-export.js gets your conversations in minutes instead of days.

It is a script you run, in your browser, on your account. It asks ChatGPT’s web app for your conversations the same way the page in front of you already does, and saves them as a conversations.json file. You then put that file in your inbox like any other export.

Why this does not compromise Mind Archive

Mind Archive never runs it, never sees your session, and never contacts OpenAI. It only ever reads a file you put in a folder. That separation is deliberate, and it is the reason this is acceptable at all — see decision D-024.

Other tools ask you to paste your ChatGPT session token into their interface. Do not do that, with any tool, including one claiming to be ours. That token can read everything in your account and send messages as you. This script never asks for it: the browser already holds the session, and the token never leaves the tab.

How to run it

Open chatgpt.com and make sure you are logged in.
Open the developer console — F12, or Ctrl+Shift+J (Cmd+Option+J on a Mac).
If the console asks you to type allow pasting, do that first. It is a safety feature, and it is right to make you think.
Open scripts/browser/chatgpt-export.js, read it, then paste the whole file and press Enter.
Wait. It prints progress and goes deliberately slowly.
conversations.json downloads when it finishes.
Move that file into data/inbox/.

What you should know before using it

The endpoints are undocumented. They are not part of OpenAI’s public API and may change, break, or be blocked at any time. If the script stops working, that is why, and the official export is always there.
It may violate OpenAI’s Terms of Use. The counter-argument — that this is your own data, which GDPR Article 20 gives you a right to port — is a real one, but it is yours to weigh. We are not going to pretend the question does not exist.
It cannot fetch images or file attachments. The official export can. If you want a complete archive, use the official export; if you want your conversations today, use this.
Going too fast can rate-limit your account. The script pauses between requests for that reason. Do not remove the pauses.

Use both

They are not alternatives. Run the script today so you can start using your archive, and let the official export arrive in its own time. Re-importing costs nothing: Mind Archive compares what is already in your archive and reports “12 new, 8 updated, 392 already in your archive”, writing only what actually changed.

Already have exports somewhere else?

Point the inbox at the folder you keep them in:

# in .env
MIND_ARCHIVE_INBOX_DIR=~/Documents/ai-exports

Mind Archive reads only .zip and .json files there, and leaves your files exactly where they are — moving things out of a folder you use for other purposes would be presumptuous. A small .mindarchive-imported.json ledger records what has already been read so nothing is imported twice.

The default, data/inbox/, is a folder Mind Archive owns. There it does tidy up: imported files move to imported/, unreadable ones to failed/ with a note saying why. An empty inbox means everything is in.

Other providers

Only ChatGPT is supported today. Claude, Gemini and others are Milestone 5 — see ROADMAP.md.