Shoot with Meta smart glasses or your phone. Every photo is auto-described on arrival, and an AI agent can search, fetch and reason over your library — by voice, in chat. Photos stay on your phone until the agent asks for one.
Get access See the pipelineA duplex loop between your phone and a stateless vision agent. Search is free; the agent pays per full-resolution photo it actually looks at — metered server-side, prepaid only.
Glasses or camera → camera roll. The observer tags the source, reads EXIF, and gets each new photo described for search. Library and chat live in local SQLite — the phone is the source of truth.
One WebSocket per turn. The agent searches your library by caption (free), pulls the one photo it needs at full resolution (one credit), answers in text and speech, and the channel closes.
5 photos/day included. Past that, paid accounts draw $1 prepaid credits — bought in advance, never billed after use. The counter lives server-side; the app can show it but can't forge it.
Sign in with Google, pick a lane. Beta is invite-only; paid is open. Both include 5 photos a day — the difference is what happens past photo six.