Cadence is the mission-critical daemon that executes your scheduled commands with zero human intervention. Boot once. Run forever. Ship every minute.
Every feature maps to a real execution primitive. No wrappers. No abstractions. Pure scheduling infrastructure.
Cadence boots at system init and immediately begins minute-by-minute evaluation of all registered task manifests. Every job is checked, every interval honored. This is autonomous execution at its most fundamental.
cadence
Per-user schedule namespacing ensures complete task isolation. Each user's scheduled workloads run in their own execution context with independent environment variables, SHELL, HOME, and PATH configuration.
/var/spool/cadence
Five pre-configured execution directories β hourly, daily, weekly, monthly, and the system-level
/etc/cadence.d/ β give you granular control over task cadence without writing a single
scheduling expression.
cadence.{hourly,daily,weekly,monthly}
Powered by inotify, Cadence detects schedule modifications the instant they hit disk. No restarts. No polling delays. Your scheduling changes propagate at filesystem speed. Falls back to modtime checks when inotify is unavailable.
cadence -i
Every job's output is automatically captured and delivered via the MAILTO environment variable.
Configure custom mail transports with -m, or disable notifications entirely. Full
observability, zero configuration.
MAILTO / cadence -m
Network-mounted shared spool directories with hostname-based leader election. Only one node in the cluster executes jobs at any given time β zero coordination overhead, zero split-brain scenarios.
cadence -c
Daylight Saving Time shifts? Cadence handles them automatically. Forward jumps trigger immediate catch-up execution. Backward shifts prevent duplicate runs. Changes over three hours are treated as clock corrections.
DST-aware scheduling
Enterprise-grade identity verification through PAM integration. Every scheduled task runs through
your existing access control policies. Configurable per-daemon with
/etc/pam.d/crond.
PAM integration
The Cadence execution timeline gives you minute-level visibility into every scheduled operation across all directories.
Cadence is designed to be zero-config by default and infinitely configurable when you need it.
Cadence starts at boot via systemd or SysVinit. It returns immediately β no background forking required. The daemon is alive and listening.
Scans /var/spool/cadence for per-user schedules, loads /etc/cadence.conf, and
indexes all files in /etc/cadence.d/. Every manifest is parsed and held in memory.
Every 60 seconds, Cadence evaluates all registered jobs against the current timestamp. Matching jobs are queued for immediate execution. Modified schedules are hot-reloaded via inotify.
Commands execute in the user's environment context with configured SHELL, HOME, and PATH. Output is captured and routed to MAILTO or syslog. Failures are logged, retries are scheduled.
Cadence's filesystem-native architecture means your scheduling topology is defined by directory structure, not YAML manifests.
Real commands. Real output. Real-time task orchestration.
Legacy schedulers weren't built for the demands of modern infrastructure.
| Capability | Cadence | Competitor A | Competitor B |
|---|---|---|---|
| Boot-time initialization | β Instant | β οΈ Manual start | β Requires orchestrator |
| Sub-minute config detection | β inotify + modtime | β Manual reload | β οΈ 5-min polling |
| Multi-tenant isolation | β Per-user spool | β οΈ Shared config | β οΈ Namespace workarounds |
| Cluster-native scheduling | β
Built-in -c flag |
β Not supported | β οΈ Requires plugin |
| DST anomaly handling | β Automatic | β Manual | β Silent failures |
| PAM integration | β Native | β | β |
| Missed job recovery | β Anacron integration | β Lost forever | β οΈ Limited |
| Pricing | From $0/mo | $$$ | $$$$ |
"We replaced three different scheduling tools with Cadence. The per-user schedule isolation alone saved us from two compliance audit findings. The inotify-based config detection means we never restart the daemon. It just works."
"Cadence's cluster mode is the single highest-leverage feature we've adopted. One flag β
-c β and our entire fleet coordinates task execution across twelve nodes without a
single line of custom code."
"The MAILTO variable is our north star for observability. Every failed job, every anomalous output β piped directly to Slack via our mail-to-webhook bridge. Cadence is the silent backbone of our entire SRE practice."
"I'm a solo founder running six microservices. Cadence's cadence.daily and cadence.hourly directories are the entire orchestration layer. No Kubernetes. No Airflow. Just directories and shell scripts. Default alive."
From solo operators to global infrastructure teams. Every plan includes the core daemon.
For individuals exploring task automation.
-s)For teams running production workloads.
-m)-n)Multi-node scheduling for growing organizations.
-c)Mission-critical scheduling for regulated industries.
-x)Cadence is designed for boot-level resilience. It starts automatically via systemd or SysVinit, loads all schedules from disk, and evaluates pending jobs within 60 seconds. Combined with anacron integration for missed job recovery, no scheduled work is ever lost.
Never. Cadence uses inotify to detect file modifications in real time. On systems without inotify, it falls back to modtime checking every 60 seconds. Either way, your changes are picked up automatically. Zero downtime. Zero restarts.
Cadence's Temporal Anomaly Resolution handles DST transitions automatically. Forward jumps (spring) trigger immediate execution of skipped jobs. Backward shifts (fall) prevent duplicate runs. Time changes exceeding three hours are treated as clock corrections and the new time is applied immediately.
Yes β Cadence's cluster mode (-c) enables network-mounted shared spool
directories with hostname-based leader election. Only the designated host executes jobs,
preventing duplicate execution. Ensure clocks are synchronized via NTP across all nodes for
optimal results.
Absolutely. Schedule files must be owned by the respective user and cannot be world-writable
or executable by others. PAM integration enforces your existing access control policies. The
-s flag routes all output to syslog for centralized, auditable logging.
Enterprise customers get additional compliance reporting and encryption at rest.
Stop babysitting scripts. Start shipping cadence. Join the teams that trust their scheduling to the daemon that never sleeps.