SmaranQ
v0.1.3 · Public beta · Available on PyPI

Camoscope

Know what your screen share may not show.

An opt-in macOS audit CLI for investigating app windows that may be concealed from a screen share. Review WindowServer sharing state, owning-process and code-signing metadata, and — when you explicitly enable Accessibility — text exposed by the app.

Camoscope

Camoscope

Check what a candidate's screen share may leave out

Check it out on Product Hunt →
Install

python3 -m pip install camoscope==0.1.3

How it works

Detect hidden-from-capture windows

Camoscope inspects macOS WindowServer sharing state and flags windows worth reviewing. The signal is behavioral rather than a simple product-name blocklist.

Verify the real process

For flagged windows, Camoscope can show the owning process and available application/code-signing metadata to help a human understand what is running.

Inspect hidden app text

With Accessibility permission, Camoscope can optionally inspect text exposed through the macOS Accessibility tree. Installing or importing Camoscope does not perform this scan.

Not just a blocklist

Because Camoscope surfaces system signals rather than issuing a cheating verdict, the output is intended as a starting point for investigation — not proof of screen-share visibility or misconduct.

Interview integrity

Use Camoscope for quick, consent-based interview monitoring.

For technical interviews or assessments, a participant can run Camoscope locally on their own Mac, with their knowledge and consent, and review or share the result. The project grew from exploring the defensive side of screen-share-resistant AI overlays while building SmaranQ, and the approach has been tested against real macOS overlay behavior including Cluely.

A flagged result is a reason to investigate, not proof of cheating or even proof that a particular capture method hides the window. A clean scan does not prove that every window or tab is visible. Camoscope cannot remotely inspect another person’s Mac.

Why I built it

The defensive side of invisible AI.

While building SmaranQ, I explored how real-time macOS assistants can stay useful during a call without becoming part of the shared screen. That raised the inverse question: if an assistant can be visible locally but absent from a screen share, what can an interviewer or proctor verify on the participant’s Mac?

Camoscope is that defensive experiment. Similar products such as Cluely and Parakeet illustrate why a screen share alone should not be treated as a complete view of a participant’s desktop. Camoscope has been tested against Cluely overlay behavior on macOS, but it does not claim guaranteed detection of any named product or version.

Consent & scope

Camoscope runs locally and is intended for transparent, consent-based auditing. It does not silently scan when installed or imported, and it does not send scan results to a Camoscope server. Accessibility remains an explicit macOS permission boundary. The project is released under MIT.