deadcron

A local dead-man's-switch for cron jobs. Cron never tells you when a job stops running — the backup that silently hasn't fired in three days, the sync that's been crashing since Tuesday. deadcron notices the silence and alerts you. Zero dependencies, no server to host.

# Replace your crontab command with a wrapped version:
*/30 * * * *  npx deadcron run backup --every 1h -- /usr/local/bin/backup.sh

# Then let the watcher check for silence:
npx deadcron install --every 5m

Why

"Cron jobs do not tell you when they fail. That is the core issue."

The classic incident: a database backup script dies at 2 a.m., keeps "running" in cron's eyes, and nobody finds out until the day you actually need the backup. Hosted monitors (healthchecks.io, Cronitor, Dead Man's Snitch) solve this — but they need an account and send your job metadata to a third party. deadcron runs entirely on your machine: a small state file plus a watcher you schedule yourself.

How it works

1. Each run checks in. Either wrap the command with run (it pings on exit 0, records the exit code on failure) or call ping <name> at the end of your script.
2. A job declares its expected period with --every (plus optional --grace). If the time since the last successful check-in exceeds every + grace, the job is overdue.
3. The watcher runs check on its own schedule (you add it to cron once via install). When a job is overdue or its last run failed, it fires every alert channel you've enabled.

Tracking a job

# Recommended: wrap the command. Auto-registers, pings on success,
# records the exit code on failure.
deadcron run backup --every 1d --grace 1h -- /opt/backup.sh

# Or ping manually at the end of a script:
deadcron register backup --every 1d --grace 1h
deadcron ping backup        # add to the tail of your job

# Inspect health any time:
deadcron status
#  ● backup    ok                every 1d   last: 3h ago
#  ● sync      OVERDUE by 2h     every 1h   last: 3h ago

The watcher

deadcron check          # evaluate now; exit 1 if anything is overdue/failed
deadcron check --json   # machine-readable
deadcron install --every 5m   # add `check` to your crontab (idempotent)
deadcron uninstall

check throttles repeat alerts (default: at most once per hour per job) so a job that's been down for hours doesn't spam you every tick.

Alert channels

check fires every enabled channel at once. Terminal is on by default.

deadcron config show
deadcron config enable macos          # native macOS notification
deadcron config set-webhook https://hooks.slack.com/services/XXX
deadcron config set-email --to ops@you.com --sendmail
deadcron config set-email --to ops@you.com --from cron@you.com \
    --smtp-host smtp.gmail.com --smtp-port 465 \
    --smtp-user me@gmail.com --smtp-pass "app-password"
deadcron config test                  # send a sample alert through all of them

Channel	How
terminal	writes the alert to stderr (great for CI / piping)
macOS	native banner via osascript
webhook	POST JSON {event, checkedAt, jobs} — Slack/Discord/your endpoint
email	local sendmail, or direct SMTP (TLS 465, optional AUTH LOGIN)

Managing jobs

deadcron pause <name>     # stop alerting (maintenance window)
deadcron resume <name>
deadcron fail <name>      # manually mark the last run as failed
deadcron rm <name>

Storage

State and config are plain JSON under ~/.deadcron/ (override with $DEADCRON_HOME). Nothing leaves your machine unless you configure a webhook or email channel.

Exit codes

Code	Meaning
0	all jobs healthy
1	one or more jobs overdue or failed
2	error (bad args, unknown job, crontab failure)

Duration format

30s · 5m · 2h · 1d · 1w (a bare number means seconds).

License

MIT