Rebuild a wearable companion app with a self-hosted backend

A Redditor who dropped their Whoop subscription had Claude Opus 5.5 rebuild the iOS and Android apps plus a private backend in about 60 hours, worth roughly $800 in API tokens.

Done withClaude· Oct 27, 2026

What they did

The poster cancelled their Whoop membership but still had the strap, and wanted to keep their health data private. They tried an existing open alternative, missed the original app's look, and instead had Claude Opus 5.5 write a full copy of it for iOS and Android, working only from screenshots and screen recordings of the official app. The apps send readings to the poster's own web service, which retrains the metric models every night. Beyond the core health and stress views, the build covers automatic activity tracking, a logbook, fitness plans, a homemade fitness-age score and a federated community tab meant to let anyone host a backend in a Docker container.

How it went

About 60 hours of work in total: roughly 23.5 hours on the iOS app, 18 on model tuning, 10 on the backend and 9 on the Android app. The poster puts usage at about 2.4 billion tokens, which Claude estimated at around $800 at API prices, mostly cache reads, all covered by a Max 5x plan. Readings like respiratory rate, blood oxygen, resting heart rate, HRV and skin temperature land within about 5% of the official app, and the fitness-age score within 8 to 10%. Everything except the strength trainer is done. It works with the MG, 4.0 and 5.0 straps.

Worth knowing

The poster will not publish the code, so none of this can be checked, and the demo video shows a simulator with fake data. The post names the model and plan but not the harness; it was posted to r/ClaudeCode. Past history can come from a Whoop data export, and the poster says the app can run fully offline with sync limited to the home network.

Try it

With Claude· 5 parts to fill in

I own a [wearable device] and want my own companion app for it instead of the vendor's. Here are [screenshots and screen recordings] of the official app. Plan the work as separate pieces: the [iOS and/or Android] app, a backend I can self-host in [Docker or my hosting], and the code that turns raw sensor data into the metrics I care about: [list of metrics]. Build one piece at a time and show me each screen in the simulator next to the matching screenshot before moving on. Keep all data on my own server. Stop and ask me before adding any feature that needs a third-party service.

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

Where they said it

Related