MQTT integration
MQTT is what your automation already listens to. A few cameras can publish to it; most can't, and none of them publish a hosted snapshot or what was in the frame. NVR Bridge does that part. Every event is published to a topic on your own broker — the language Home Assistant, Node-RED and openHAB already speak — so whatever you've already automated can act on it.
topic
nvrbridge/back-gate-cam/event
{
"camera": "Back gate",
"eventTime": "21:04:11Z",
"objects": [
{ "label": "person", "score": 0.94 }
],
"image": "https://…/optical.jpg"
}How it works
Each camera gets its own private inbound address. Point the camera's existing email or FTP alert settings at it — no firmware changes and nothing to install.
Add an MQTT notification with the broker's host, port and credentials, then choose the topic, the QoS and whether to retain. If the broker only exists on your LAN, give us SSH details instead and we'll publish through a tunnel — nothing is exposed to the internet.
A trigger describes what matters — motion on the back gate, a person on the driveway, one particular number plate. Attach the MQTT notification to it and only those events go out. One trigger can feed several destinations at once.
Hit Test to push a sample event with sample detections, so you can see exactly how it lands in MQTT before a real one arrives. Every delivery after that is logged with its result.
What you get
Questions
No. Give us an SSH host and key and we'll open a tunnel for the moment it takes to publish, then close it. The broker keeps listening on your LAN only, exactly as it does now.
JSON by default — camera, time, event type, snapshot links and any detections. If the receiver wants a particular shape, write the body yourself with placeholders.
No. The camera connects outward to us using the email or FTP settings it already has. There's no recorder, no mini-PC, no VPN and no inbound port to open.
Yes. Turn on detection and filter the trigger to what you care about — a person, a vehicle, a particular animal, any number plate or one specific plate. Everything else still lands on your timeline; it just doesn't interrupt you.
Add a camera, point its email at the address we give you, and watch the first event land. About five minutes, all of it in a browser.