Private AI memory app for Mac. It records your screen and calls (Zoom, Meet, Teams, Slack Huddles) locally, OCRs everything on-device using Apple’s Vision framework, and indexes it in a local SQLite database with full-text and vector search. Ask questions about anything you saw or heard, and get answers grounded in your own history. It supports Bring Your Own LLM via any OpenAI Responses API endpoint—including a model running on localhost, so even inference stays on your machine—plus optional encrypted cloud backup to an S3 bucket you control, and MCP integration. Free to download and run locally; paid plans add capacity and cloud storage. macOS only.
Pros & Cons:
Pros: Local-first with clear privacy boundaries; on-device OCR and local indexing; BYOLLM lets even inference stay on your machine; free to download and run locally; optional encrypted backup to an S3 bucket you control; unusually candid disclosure of network-call boundaries.
Cons: macOS only, with no mobile or Windows build; not open source; paid pricing not publicly listed; the Teams product line is separate and not covered by the personal version; requires configuring an LLM endpoint, which adds some setup friction.
Who is it for?:Privacy-conscious Mac users who care about data sovereignty; individual knowledge workers who need “screen/call memory search”; technical users willing to BYOLLM.