Your sources, your choice
The intended desktop app indexes selected screenshots and images. It does not require continuous screen recording, automatic file relocation, or whole-device access. Capture integrations require their own transparent permissions.
Local processing by default
The design target is on-device OCR, embeddings, notes, and search after required model assets are installed. Offline behavior and actual network activity must be verified before a desktop release.
Derived data matters
Notes, OCR, thumbnails, vectors, file paths, and source URLs can reveal information about originals. Storage protection, deletion, export, and revocation need to cover these derived records too. An encrypted database or completed security audit is not claimed today.
Assistant access is a separate boundary
Enable the connection deliberately and limit searchable sources. A cloud assistant receives the content returned to it. Local search plus selective sharing is different from fully local inference. Disabling access should stop future requests, but cannot recall content already shared.
Moving to another laptop
Portable library archives and optional passphrase protection are planned. Complete exports include selected originals; context-only exports need relinking. Assistant credentials and machine-specific access are not carried over. Read the transfer boundary →
The website preview
The static site searches sample records and saves edited sample notes in browser local storage. It has no file upload endpoint, AI API calls, or assistant connection. Copy and export only run when requested. Clearing browser storage removes note edits. A hosting provider may process ordinary request logs under its policies.