Build a desktop popup that shows Frigate cameras when motion is detected

A developer had Claude Code write FrigatePiP, an open-source Windows and Linux app that pops up a live camera feed when Frigate spots a person or car, and now uses it every day.

Done withClaude CodeinGitHub· Nov 19, 2026

What they did

The poster, a developer who sits at a PC most of the day, missed a fox taking shoes from outside the door and only saw it later in the recordings. They wanted the camera to come to the screen instead. They wrote the spec and made the design calls, and Claude Code wrote most of the code for FrigatePiP: a small app that listens to Frigate events over WebSocket or MQTT and opens a corner window with that camera's live stream, closing it once motion stops. It ships as a Windows tray app and for Linux desktops, framebuffers and terminals, with filters by camera, object, zone and confidence. The poster reviewed the code and approved each pull request, and tested it against their own cameras.

How it went

The app works and the poster runs it daily. They say it idles at 5 to 10 MB of RAM on Windows with near zero CPU, decodes H.264 and H.265 in hardware and falls back to MJPEG or snapshots. It made Frigate's Third Party Extensions list. A commenter running it on a Raspberry Pi 4 next to a Home Assistant dashboard said it was a big improvement on their previous setup. The framebuffer mode had not been tried on a real Pi before release.

Worth knowing

The Windows installer is unsigned, so SmartScreen warns on install. With the Home Assistant Frigate add-on, port 5000 is off by default and must be enabled, or use 8971 with a viewer user. It needs Frigate for detections, so a Reolink NVR alone will not work.

Try it

With Claude Code· 5 parts to fill in

Build a small desktop app for [Windows, Linux or both] that listens to events from [Frigate or my NVR] over [WebSocket or MQTT]. When an event matches [cameras, objects, zones, minimum confidence], open a small always-on-top window with that camera's live stream from [RTSP or go2rtc URL], and close it when the event ends. Use hardware decoding where available and fall back to snapshots. Keep idle CPU and memory low. Work in small pull requests and stop for my review before each merge.

Square brackets are the parts to fill in. The prompt is ours; the story is theirs.

Where they said it

Related