COMPANY — ABOUT US

About VisionGuard

VisionGuard turns the cameras a site already has into plain answers — who arrived, how busy an area is, where someone went — instead of hours of footage nobody has time to watch. It's built in Egypt for offices, factories, warehouses and hospitals that run their own cameras and want a straight answer, not another screen to babysit.

Who it's for

VisionGuard is for the person who already has a wall of cameras and not enough hours to watch them: an office manager checking who's in, a factory floor keeping a restricted area clear, a warehouse watching who's near the loading bay, a hospital tracking who came and went. You don't replace your cameras or your team — you point VisionGuard at the feeds you already have and ask it what's happening.

Honest by design

A camera system that guesses, or hides what it doesn't know, is worse than useless in a building full of people. VisionGuard is built around rules it does not break:

  • A face names a person. Clothing only ever says "same body" — the two are never merged into one result. Green means confirmed by face, blue means matched by appearance, amber means not yet identified.
  • If a camera, a model, or a connected service is offline, the screen says so, with the reason — never a stand-in answer.
  • Sample data is always labelled "demo." It only appears when you ask for it.
  • Every knife alert goes out marked unconfirmed until a person reviews it, and the entry — confirmed or false alarm — is never deleted from the history. The video behind it is kept for 30 days.
  • Searching for a person, or asking a question, is logged before you see the result — who looked, and when.

Built to run at your site

VisionGuard runs on a computer at your own site, not as a cloud service watching your cameras from somewhere else. Camera footage and faces never leave the building. Add the cameras you already have by their address — nothing new to install at the camera end.

Two things do leave the site. Alert emails go out through Gmail and carry facts only — what, where, when, how sure, who decided — never a name, a face or a picture. And when you type a question, the words — with names and plates masked into placeholders first — go to Google's Gemini, on its free tier, to work out what you're asking. Google's terms let them use what's submitted and let a human reviewer read it; we accepted that because masking cuts what leaves down to a sentence like "when did PERSON_1 arrive today?" A word the system doesn't recognise — a visitor's name is the usual case — is sent exactly as typed.

Talk to us

Want to see VisionGuard against your own cameras? Get in touch.

Last updated 27 September 2026