Import conversations from ChatGPT, Claude, Claude Code, Gemini, and Grok into one local archive. Search across everything. Continue old threads. Keep the work on your machine.
Core archive features stay local. Optional intelligence only contacts the LLM provider you configure and trigger.
Works with
One local archive for every AI conversation you've ever had.
Drop your ChatGPT .zip, Claude .json, Claude Code .jsonl, Gemini Takeout, or Grok .json. Provider auto-detected. Duplicates skipped.
SQLite FTS5 across all providers at once. Snippet highlights and deep links back to the exact message in the transcript explorer.
Prompt-level table of contents, minimap rail, and full threading. Like reading a real archive.
Optional grounded answering over your archive, with citations. Uses the LLM provider you configure; returns "insufficient evidence" instead of guessing.
Distill, Continuity Packets, and bridge handoffs turn a selection of conversations into paste-ready context for the next chat.
Markdown, attachment-aware zip bundles, an Obsidian raw-inbox bridge, and a verifiable Memory Passport with manifest, JSONL lanes, and checksums.
Clone, install, run. Your conversations appear instantly.
Download your conversation history from ChatGPT, Claude, Claude Code, Gemini, or Grok. Each has an export option in settings or session files.
Drop the file on the Import page. Provider is auto-detected. Normalized into one canonical SQLite ledger.
Full transcript explorer with TOC and minimap. Cross-provider search returns results with provenance.
Export a verified Memory Passport. Your archive is one SQLite file: portable, inspectable, yours.
Open source. Local-first. Apache-2.0 licensed.
No accounts, no analytics, no telemetry, no background sync.
Older packaged builds are on GitHub Releases, but source install has the newest fixes.