How a screen-aware AI assistant works

"Screen-aware" means the assistant sees your display as input — the code, the deck, the dashboard — so your questions don't have to carry all the context in the prompt.

What actually happens

Frames from your screen (or a browser share) go to a vision-capable model along with your question. The model answers with both as context — so 'is this right?' about a wiring photo or a stack trace just works.

Requests route through Cloudflare AI Gateway, which logs usage and enforces spend controls on our side. Your screen content isn't stored as recordings.

What it can and can't do

It reads what's visible — text, code, diagrams, photos of physical work. It can't see windows you haven't shared, and it doesn't retain frames between sessions. If it's unsure about what it's looking at, the HUD shows that instead of bluffing.

Why we built it this way

We started this after a chatbot confidently told us a battery hookup was fine. It wasn't — and a text-only assistant had no way to catch that. An assistant that can see the actual work is the difference between advice and a guess.

The full story is on the landing page.

Try it