An AI chat backup you actually own

Your AI history is split across providers — ChatGPT chats in an OpenAI account, Claude chats in an Anthropic one — with no copy that's yours. Mirror turns the official export from each into a single encrypted archive you can search and keep.

One timing difference worth knowing: Claude's export usually lands in minutes; ChatGPT's often takes hours, sometimes a few days. Start with whichever arrives first — or don't wait at all.

1 Request the export2 Upload it to Mirror3 Browse your archive

Free, no card — email sign-in only when you're ready to upload.

Start today — don't wait on the ChatGPT export

Two of these paths work in this session; the third is worth kicking off now so it's ready when it lands.

Claude export — usually minutes

Request the export in Claude's privacy settings — the download link typically arrives within minutes. Drop the .zip in as it arrived for a working archive this same session. Save Claude conversations guide.

After you upload

Once your export is in Mirror, ask your chats in plain language — or connect your agents so Cursor / Claude Desktop share one memory.

ChatGPT Takeout — request now, merge later

ChatGPT's official export can take days — request it now. When the email lands, drop the .zip in as it arrived; big zips are stripped down to your chats before upload. Export ChatGPT chats guide.

Why Claude first, ChatGPT second

Claude's official export usually lands in minutes — often the same session — so start there when you want a working archive today. ChatGPT Takeout can take days, so request it now and merge when the email lands. Soft Pro Capture keep-current recruiting stays on hold.

How Mirror works

1 Official export

Request the data export in your provider's settings. They email you a .zip — no scraping, no API keys, nothing extra to install.

2 Upload once

Drop the official .zip in as it arrived — whatever size it is. Mirror strips it down to conversation data in your browser before upload, then it's encrypted and tied to your account.

3 Browse your archive

Mirror counts the conversations, messages, and words that exist only in the provider's account — and surfaces the threads you'll actually recognize.

Get your export

ChatGPT can take days

  1. Sign in at chatgpt.com, open your profile menu, and choose Settings → Data controls.
  2. Under Export data, click Export, then Confirm export.
  3. OpenAI emails a download link — often hours, up to a few days for large accounts — and it expires 24 hours after it arrives.
  4. Download the .zip while signed in, then drop it into Mirror as it arrived — big Takeouts are fine; we strip them down to your chats before upload.

Claude usually minutes

  1. Sign in at claude.ai, click your initials, and choose Settings → Privacy.
  2. Click Export data and confirm.
  3. Anthropic emails a download link — usually within minutes; it expires 24 hours after delivery.
  4. Download the .zip while signed in, then drop it into Mirror as it arrived — we prepare it for upload automatically.

Timing, size limits, and common mistakes: full step-by-step export guide.

Frequently asked questions

How long does ChatGPT vs Claude export take?

Claude's official export usually lands within minutes. ChatGPT Takeout often takes hours and can take a few days — it is not instant. Request ChatGPT now, start with Claude today, and drop the Takeout zip in as it arrived whenever it lands — big zips are prepared automatically before upload.

What does Mirror back up?

Mirror builds an encrypted archive from the official data exports for ChatGPT and Claude only — no scraping and no other providers on the door. Your exports are never used to train models.

How do I keep the backup current after the first upload?

Re-export periodically and drop new zips into Mirror. Then ask your chats or connect your agents so the vault stays useful. Soft Pro Capture keep-current recruiting stays on hold.

After you upload: Once your export is in Mirror, ask your chats in plain language — or connect Cursor / Claude Desktop so your agents share one memory.

Mirror works with the official data exports from ChatGPT and Claude only — no scraping, no other providers. Your exports are never used to train models. Soft Pro Capture keep-current recruiting stays on hold.

Start free — add your export