Importing ChatGPT and Claude History into MemPalace — Step by Step

What You Will Need

To complete this guide you need: MemPalace installed, a palace initialised (the mempalace init step creates your palace structure), and access to your AI conversation exports. The export process is slightly different for each platform, but takes under five minutes in each case.

The Five Supported Conversation Formats

MemPalace's conversation miner (convo_miner.py) automatically detects and normalises five different export formats into a consistent internal transcript structure before ingestion:

Exporting Your ChatGPT Conversation History

1. Sign into ChatGPT and click your profile picture in the top-right corner
2. Select Settings from the dropdown menu
3. Navigate to the Data Controls tab
4. Click Export data, then confirm by clicking Confirm export
5. ChatGPT sends a download link to your account email — usually within a few minutes
6. Download and unzip the archive. Your conversation data is in conversations.json.
7. Move the file to a dedicated folder, for example ~/chats/chatgpt/conversations.json

Exporting Your Claude Conversation History

Claude.ai Projects

1. Open Claude.ai and navigate to the project you want to export
2. Look for an Export or Download option in the project menu or settings panel
3. Download the resulting JSON file and place it in ~/chats/claude/

Claude Code Sessions

Claude Code saves sessions automatically to ~/.claude/projects/ as JSONL files. No manual export is required — you can mine directly from that directory:

Splitting Large or Concatenated Export Files

Some platforms — and some export tools — combine multiple conversation sessions into a single large file. MemPalace's split command breaks these into individual per-session files, which produces cleaner room assignments during mining:

# Check what the splitter would do without changing any files
mempalace split ~/chats/ --dry-run
# Run the actual split
mempalace split ~/chats/
# Only split files that contain 3 or more sessions
mempalace split ~/chats/ --min-sessions 3

Mining Your Conversations into the Palace

With your files split and ready, run the mining command. Choose the mode that matches your goal:

# Standard mining — stores everything verbatim
mempalace mine ~/chats/ --mode convos
# Assign all mined content to a specific wing
mempalace mine ~/chats/ --mode convos --wing myproject
# Auto-classify content into hall types (decisions, milestones, preferences...)
mempalace mine ~/chats/ --mode convos --extract general

The --extract general flag runs auto-classification on top of standard mining. It takes longer but organises content into the five hall types: hall_facts (locked decisions), hall_events (sessions and milestones), hall_discoveries (new insights), hall_preferences (habits and opinions), and hall_advice (recommendations).

Verifying the Import

mempalace status
→ Wings: 4  |  Rooms: 27  |  Drawers: 2,341  |  KG triples: 89
# Test with something you know you discussed
mempalace search "why did we choose this database"
→ Returns verbatim passages from your actual conversations

If the search returns relevant passages you recognise, the import worked correctly. If the palace shows zero drawers, check that your chat export file was found — run the mining command with the explicit file path to confirm.

Importing Slack Exports

1. In Slack, open your workspace settings and navigate to Import and Export Data
2. Click Export, then select All Messages from the date range options
3. Slack will email you a download link for a ZIP archive containing JSON files organised by channel
4. Unzip the archive and mine the resulting directory:

Assigning a wing name with --wing keeps Slack content separated from your personal AI conversations, which makes retrieval more precise.