AI agents for physical security
Novin

The intelligence layer behind the cameras.

Cameras record. Detectors label. Neither understands the place. Novin is the intelligence layer behind the cameras: vision that reads the scene, memory that never resets, and reasoning that works like a sharp human team. You understand what happened without opening the clip. Built for explainability. That is what we sell.

Request access See products
Vision Reads the scene. Who, how they move, what a person would notice.
Memory Infinite. This site, this hour, this camera, kept.
Reasoning A call you understand without watching the footage.
The team at work Working the event
Commercial perimeter Loading bay side gate 02:17 AM
1 / 3

Vision, memory, and reasoning on one event.

The team looks, remembers, and reasons. Then it tells you what happened, so you do not have to watch.

Vision
Read the scene.

Not a motion blob. Who is there, how they move, what they are doing. What a person would notice first in the image, clip, or snapshots.

Memory
Remember this site.

Infinite memory of this place. This hour. This camera. Site context: what usually happens here.

Reason
Reason like a crew.

Does this belong at this hour, on this camera? Should a human be woken? They talk it out, then sign the call with a trail.

Call
Make the call.

A decision you can understand without looking. The trail is the product, not a log on the side.

Result: you know what happened. You do not have to open the clip. Explainable

What the team noticed

Observed movementControlled
Time signalUnusual
Site confidenceMixed

Recall context

Prior behaviourPartly known
Route expectationBroken
Decision trailReviewable
Final call: keep the event visible, keep the trail, and route it the way a duty manager would. Team call
Vision Reads the frame the way a person would. Behaviour, objects, path. Not a motion flag.
Reasoning Weighs the scene against the site. A call, not a score. With a trail.
Memory Infinite memory of the site. Context compounds. The team gets sharper on this place.
Explainable You understand what is happening without having to look. Built for that.
The strengths

Vision and reasoning, in one layer behind the cameras.

We sell the intelligence layer. Vision, reasoning, infinite memory, and explainability. You know what happened without watching the footage. Configurable from the CLI.

Vision that actually reads the scene.

Not motion. Not a person-box. The team looks at the image, the clip, or a burst of snapshots and sees what a person would see: who is there, how they are moving, what they are doing.

Result: vision with judgment already attached.

Reasoning that works like a crew.

They talk it through. Does this belong at this hour, on this camera, at this site? Should a human be woken? You get the story of the event, not a score you have to go and check.

Result: you understand the event without looking.

Infinite memory. Site context.

The team does not start from scratch. This building, this garden, this loading bay, kept. Context is built in, not typed in each time.

Result: a team that knows the place it watches.

Built for explainability.

The words are enough. You should not have to scrub the tape to know what happened. Image, video, or snapshots still come in. The explanation is what you act on.

Result: a decision you understand. CLI when you want to shape the team.

Where it sits

Same layer. Behind commercial cameras and home cameras.

One product. Vision plus reasoning plus memory, behind the cameras you already run. On a site or in a home, the team remembers that place.

Novin
Commercial

Behind the cameras on a commercial site. Vision reads the leftover person. Reasoning decides if a duty manager should see it. Memory knows this loading bay at 02:17.

How the team sounds
Activity happens at an unusual timeRisk level increases
Behavior matches a real threatEscalate immediately
Operators receive a clear explanationFaster response
  • Vision plus site context, so leftover tickets and after-hours movement get a reasoned call.
  • Explainable: a duty manager knows what happened without scrubbing the tape.
  • Send image, video, or snapshots from the stack you already have.
  • Exportable reports for enterprise audits and customer reviews.
Novin Home
Residential

Behind the cameras on a home. Vision sees kids, pets, a delivery, the wind. Reasoning knows that is life. Memory keeps the household.

How the team sounds
Matches family's normal routineIgnore the alert
Recognizes expected behaviourRoutine protected
Only alerts on real anomaliesTrust maintained
  • The layer sits behind the doorbell and the garden cam. The household keeps the app they already have.
  • Alerts that actually mean something when they come through.
  • You understand the event without opening the clip. That is the product.
  • Privacy-first approach keeps things local and practical.
How It Works

Send. See. Remember. Reason.

You send an image, a video, or snapshots. Vision reads the scene. Memory checks the site. Reasoning makes the call. You understand it without watching.

1. Send

Hand over an image, a video, or a set of snapshots from the cameras you already have.

Image Video Snapshots
2. See

Vision reads the scene. People, behaviour, objects, path. What matters in the frame, the way a person would see it.

Vision Scene
3. Remember

Infinite memory and site context. Has this happened here, at this hour, on this camera?

Infinite memory Site context
4. Reason

Reasoning talks it through, then tells you what happened. You should not need to look at the clip to understand it.

Reasoning Explainable
Access

Buy the intelligence layer. Put it behind your cameras.

The product is the intelligence layer: vision, reasoning, infinite memory, and explainability. You understand what happened without opening the clip. Image, video, or snapshots in.

Talk to us Join home waitlist
Vision. Reads the scene in the image, clip, or snapshots. Not a motion flag.
Reasoning. Talks it through like a crew. You know what happened without watching.
Infinite memory. The site is kept. Context compounds. The team gets sharper on this place.
CLI. Shape the team from the command line. Then let it watch.