Search, analyze & export your history

Full-text search across every message

Enter a phrase — a library name, a bug ID, an idea — and the analyzer returns matching messages with the conversation, role, and timestamp. Search covers user, assistant, and system messages alike, and runs against an in-memory index built on a Web Worker. On large archives you get instant results without the tab freezing.

Activity heatmaps & usage signals

Message counts, day-of-week and hour-of-day heatmaps, token estimates and cost approximations show how you actually use the models. Useful for spotting patterns — the 2 a.m. debug loops, the day you exported six tickets of context, or simply how much conversation you've generated.

Custom instructions & system prompts

Every instruction your assistants were given — ChatGPT custom instructions, Claude projects and system prompts — shows up in the export as system-role content. The analyzer pulls these into a dedicated panel so you can audit exactly what each model was told, spot drift between projects, and recover instructions you thought were lost.

Export to Markdown, text or JSON

The original export is optimized for machines. For humans, export to a clean, readable transcript:

• Markdown — per-conversation transcripts with headers, formatting and timestamps, ready for notes or a wiki.
• Clean text — plain, stripped audio-worthy copy.
• JSON — a normalized, re-exportable mirror of your archive.

Every export is generated locally and downloaded as a file. No watermark, no round-trip, no server.

Putting it together

Export, analyze, scan for secrets, search, then export a clean copy. The whole workflow never leaves your computer.