Glue that makes your AI cameras talk to Home Assistant

What it does

This is the official Home Assistant integration for Frigate, the open-source NVR that does local AI object detection. It exposes Frigate’s internals—camera streams, detection metrics, motion states, and recording toggles—as native Home Assistant entities you can automate against. It also adds a media browser for scrubbing through event thumbnails.

The interesting bit

The integration doesn’t just mirror cameras; it surfaces operational telemetry like Detection FPS and Skipped FPS as first-class sensors. That turns Frigate from a black-box NVR into something you can monitor and script like any other HA device. Multiple Frigate instances are supported, which matters if you’re running cameras across separate hosts.

Key highlights

• Native sensor entities for camera, detection, process, and skipped FPS
• Binary sensors for object motion, plus switches for recording, detection, snapshots, and contrast boost
• Camera entities for live view and object-detected snapshots
• Services for manual events and PTZ control
• Rich media browser with thumbnail navigation (requires media_source enabled)
• Companion Lovelace card available separately

Caveats

• Requires MQTT to be manually configured in Home Assistant first; the integration won’t set it up for you
• HACS installs the code but you still need to add the integration through HA’s UI afterward

Verdict

Worth it if you’re already running Frigate and want to wire detections into automations, dashboards, or voice assistants. Skip if you just need a camera feed in Home Assistant—there are lighter ways to do that without an NVR in the loop.