Exports
Alerts
Noticed on your cameras
Worked out by counting what each camera normally sees — no model, no cost. Accepting one creates the rule switched off, so you can look at it first.
New rule
Recent
Search footage
Describe what you are looking for in ordinary words — man in a hi-vis jacket, white van, someone on a bicycle. This ranks recorded events by how much they look like the description. It does not identify anyone, and the top result is the most likely candidate rather than a confirmed match.
AI backend
Two roles, configured separately. Compiling a rule happens once and only ever sees the sentence you typed. Adjudicating happens per event and sees a cropped image from your camera — so a site that cannot let footage leave can run that one locally while still compiling in the cloud.
These settings are managed by your provider. You can see which model is in use and what it is costing, but only they can change the backend or the spend cap.
Spend cap
At the cap, degrade pauses only the rules that need a model. Rules built from zones, timing and size keep alerting — an alerting system that goes silent is worse than one that gets simpler, because silence looks the same as nothing happening.
Usage
| Day | Backend | Calls | In | Out | Cost | Errors |
|---|
Team
Who can sign in, and what they may do. Viewer watches. Operator also exports and fetches from cameras. Admin changes cameras and rules. Owner additionally manages people.
Add someone
| Name | Role | Sites | Last signed in |
|---|
API keys
For other systems and agents. A key carries a role, so give it the least it needs — viewer can ask what happened and watch; it cannot change anything. The machine-readable description of what a key can call is at /openapi.json.
Where alerts are sent
An alert that only lands in this page is one somebody has to be watching for. A webhook pushes it to Slack, a phone, or your own system the moment it fires. Each payload is signed, so the receiver can prove it came from here.
Sites
Each site needs its own bridge. Adding one gives you a token to put in it.
Single sign-on
Let people sign in with your company's own accounts — Google Workspace, Microsoft 365, Okta or anything else that speaks OpenID Connect. Passwords keep working either way, so a provider problem can never lock you out.
Single sign-on is set up for you by your provider. You can see how your staff sign in; contact them to change it.
1. Your email domains
A domain is what tells the login page to use your provider. It has to be verified first — otherwise anyone could claim a domain and intercept its sign-ins.
| Domain | Status | TXT record to add |
|---|
2. Your provider
Register this redirect URL with your provider first — it will refuse the sign-in otherwise, and the error it gives does not say which URL it wanted:
With create accounts automatically off, someone who signs in successfully but has not been added here is turned away — which is what you want when the directory is larger than the group who should see cameras.
Customers
Creating a customer makes a tenant, its first site, and the bridge token for it. You manage their AI backend and spend cap; they cannot raise it.
| Customer | Cameras | Storage | AI spend |
|---|
Fleet & storage
Recording health, measured bitrate and storage caps. Click a value to change it.
Bridges
Find cameras on the network
Or add one by its RTSP address
For cameras a scan will not find — old ones with no ONVIF, or a version of it we cannot read. Recording, motion, classification and rules all work exactly the same; what you lose is discovery, camera-reported motion events and SD-card backfill, because those are ONVIF features.
| Camera | Device | Address | State | Measured | 30-day | Used | Cap (GB) | When full | Keep | Mode | Detect | Motion | SD card |
|---|