Cronitor

Promptfree, for everyone, and the only version there is

Open the prompt in Claude CodeOpen the prompt in CodexOpen the prompt in Cursor

Copy

Build me dead-man monitoring for scheduled jobs.

THE IDEA
- A job that fails loudly is easy. A job that silently stops running is what this catches

STACK
- A small always-on worker on a different machine from the jobs, SQLite

PINGS
- GET /p/:key/start and /p/:key/done, plus /fail with an optional message
- Each monitor has an expected schedule as a cron expression and a grace period

THE CHECK, WHICH IS THE PRODUCT
- Every minute, find monitors whose next expected finish plus grace has passed with no done ping
- Raise an alert once per incident, not once per minute
- Resolve automatically on the next successful done ping

DURATION
- Store start and done times; alert when a run exceeds a multiple of the median of the last twenty

ALERTING
- Webhook first, then email; both retried with backoff
- A daily heartbeat mail proving the monitor itself is alive, because nothing watches the watcher

Include the schema, the missed-run query and the heartbeat.