Your idea is possible.
Here’s what actually fits.

A small MIT-licensed project. Useful reference for pairing, a capture button, and streamed answers. Turning it into phone-camera → VPS needs a different capture client and authenticated internet transport.

MIT · small project

Current README calls it source-available under a personal-use, non-commercial license. Do not assume GitHub availability means unrestricted reuse. Its stealth and speed claims are publisher claims, not independently established guarantees.

Source-available · restricted reuse

The current v1 app is proprietary; the old GPL code remains in repository history. Linux capture behavior differs between X11 and Wayland. “Invisible” in marketing is not proof of OS-level invisibility.

Current v1: closed source

Start here to test whether screen text stays readable at a fixed distance. Controls depend on the camera hardware and Camera2 support. AF/AE support on a particular Realme lens still needs an on-device test.

Existing camera app

The Play listing documents local Wi-Fi viewing and optional cloud broadcasting. This could test camera-to-server ingestion. For a real deployment, use authenticated transport and avoid exposing an unprotected camera feed. The app contains ads.

Network camera · AI not included

A useful reference for browser camera input and live responses. Server authentication, phone pairing, persistent camera behavior and your chosen API need implementation. Android browser focus/exposure controls require device testing.

Apache-2.0 · developer starter

The homepage emphasizes no meeting bot and an overlay excluded from screen shares. These are narrower claims than invisibility to operating-system monitoring. Its advertised transcription timing is not a benchmark for complete image-question solving.

Commercial product

A practical starting point if you allow capture on the laptop: the upload workflow can target your own service. This removes the need to write basic Windows capture from scratch.

GPL-3.0 · Windows utility

Useful for an ordinary Ubuntu screenshot-to-server workflow. A command-line capture can reduce interaction, but running without a visible window is not the same as being unobservable.

Open source · capture utility

Better suited to ongoing desktop memory than a minimal capture button. Current documentation describes the code as source-available. Local collection and any cloud AI/sync are separate data paths to evaluate.

Source-available · desktop recorder

Useful as an interface reference, not a current turnkey recommendation. Review dependencies, model configuration and API-key handling before adapting it.

Archived · MIT

Useful for prototyping camera quality, but not a match for “nothing runs on the laptop.” Publisher mirroring-latency figures describe video transport, not GPT completion time.

Apache-2.0 · desktop receiver

A convenient webcam product when using software on the computer is acceptable. Camera-to-AI integration would still be separate.

Free core · optional upgrades

Closer to the original “screenshot the camera app” idea, but it adds an unnecessary display-capture layer. Native camera frames are the cleaner engineering path for reading a distant monitor.

Early project · license not verified

Useful inspiration for saved captures and searchable answer history. It does not replace remote capture, phone pairing or VPS orchestration.

GPL-3.0 · screenshot organizer

A similar screen-to-answer concept, but not an independent phone-camera system. It was reviewed as a public project; its detection-related claims were not audited.

GPL-3.0 · desktop app

Conceptually similar to other desktop assistants. No reviewed evidence establishes invisibility to privileged monitoring or an existing phone-camera-to-VPS mode.

MIT · claims need validation