Your AI chat history can contain details you no longer remember sharing: a salary pasted into a budget question, an address in a draft, or an API key in a debugging session.
Memory Mirror is a ufodia workbench for reviewing supported chat exports locally in your browser. It brings potentially sensitive details back into their conversation context so you can decide what deserves attention.
Try it in three steps
Open the workbench and choose Try the synthetic demo. Start with fictional data to see how the review works.
Explore the categories and findings. Read the source context, try a watch word, and review which conversations need a closer look.
Build a cleanup checklist or export a report. If something needs removing, review it in its original app.
Memory Mirror supports documented export formats, including ChatGPT, Claude, Open WebUI, and SillyTavern. It needs no account, API key, or model download. Analysis stays in your browser.
The detectors use patterns, checksums, and word lists. They can miss details or flag innocent text. This reviews exported content; it cannot reveal hidden model memory or delete account data. Check downloaded reports before sharing them.
Try Memory Mirror · Explore the MIT-licensed source
If this makes a difficult review more manageable, one coffee helps support ufodia’s testing, documentation, and improvements. Thank you.
Screenshot below: the built-in synthetic demo uses fictional data. The score is a review aid, not a measurement of risk.
