Migrating from Claude Pro to Teams? Here’s how to not lose everything.

So Anthropic finally decided I need to share my AI uncle with the rest of my team and move to a Claude Teams account. Great. Except nobody told me that when you make that switch, your entire conversation history, all your projects, and every single file you ever uploaded just… stays behind. On the old account. That you’re about to lose access to.

I had 223 conversations, 18 projects, and years worth of uploaded files ranging from logs to GPS data files, medical records, concert photography scripts, and about 5000MB of other stuff I actually need.

Anthropic does have a “Data Export” option in settings. I tried it. It’s basically useless — you get a JSON with conversation summaries and almost none of the actual content or files. So that was a dead end.

I’m not going to manually copy-paste 223 conversations. So I did what any reasonable tech person would do at 11pm — I built a scraper.

What claude-takeout does

It’s a Python tool that:

• Exports all your conversations grouped by project, with full transcripts
• Downloads all uploaded files — images, ZIPs, PDFs, EMLs, CSVs, videos, basically everything
• Builds an import package — a searchable offline HTML archive and Markdown docs you can paste straight into your new Team account’s project knowledge

The whole thing runs in a browser window (you can watch it work) and uses Claude’s own internal API to get the data properly, not just scraping the DOM hoping for the best.

How it works

Three scripts, run in order:

1. scraper.py — The main one. Opens a browser with a saved session (you log in once, it remembers), hits Claude’s internal API to list all your projects and conversations, then visits each conversation to grab transcripts and intercept any files that load. Took about 90 minutes for my 223 conversations.

2. download_zips.py — Handles the binary files (ZIPs, PDFs, EMLs etc.) that the API doesn’t serve directly. It goes through every conversation, finds the file download chips, clicks them, handles the preview modal that pops up, and saves everything. Has a retry system that goes up to 5 rounds with increasing delays between them for the stubborn ones.

3. importer.py — Takes the exported data and builds your migration package: a searchable offline HTML archive with project filters, one Markdown doc per project ready to paste as knowledge, and a master context document for your new account.

The honest bit

It works really well. Most things get exported perfectly. But I should be upfront about what doesn’t:

“Pasted” text chips — when you paste code or text directly into Claude rather than uploading a file, it shows as a chip but there’s no actual file behind it. The content is in the transcript though, so nothing is lost, it just won’t appear in the files folder.

Some very large files — occasionally Claude’s servers just time out preparing the download link even at 90 seconds. The retry system catches most of these but a handful might need manual downloading. Mine had about 38 that refused to cooperate through all 5 retry rounds, most of which were duplicate uploads or files that appeared broken even when trying manually. Unfortunately, for those files I was not able to download manually from my browser or even the Claude official app, so it’s a Claude issue, not the script’s problem.

The official data export — still useless, but you already knew that.

Setup

pip install -r requirements.txt
playwright install chromium

python scraper.py        # grab everything
python download_zips.py  # get the binary files
python importer.py       # build the import package

First run opens a browser, you log in once, session is saved. After that it runs unattended — I ran download_zips.py overnight on my home server.

The result

223 conversations captured. 18 projects preserved with their structure. Thousands of files downloaded. One searchable HTML archive that works offline. And a set of Markdown docs I can paste into my new Team account so the new Claude knows what we’ve been working on.

The tool is open source on GitHub: https://github.com/BeforeMyCompileFails/claude-takeout

If you’re making the same migration and don’t want to start from scratch — give it a go. And if you break it or make it better, PRs are very welcome.