Service Monitor keeps an eye on your websites and services. Every minute, it checks that they answer properly. When one stops answering, you get one alert, with the evidence. When it’s back, you get one “all good”. No alert storms.
It’s the first piece of a bigger plan: tools that make AI apps easier to run, and easier to fix when they break.
I’m building it in Rust with Claude Code, and writing up every step here, including the mistakes.
The rules it lives by
- No green without proof. If it didn’t check, it says “unknown”, never “fine”.
- One check, one try. No hidden retries that hide problems or pile onto a struggling service.
- Everything has a limit: time, retries, memory and disk.
- Written in Rust, where every piece of data has exactly one owner and things clean up after themselves.
- No bragging without numbers. No speed or memory claims until a 24-hour test backs them up.
Early days: nothing here is production-ready yet.