Schedule server jobs

Automations run on the agent. They can execute a shell command, send an HTTP request, or send a game-console command. They are not a cloud scheduler that runs jobs while the agent is stopped.

Before you begin

You need access to the server and permission to manage its automations. Shell and game-console actions require the corresponding agent capabilities. Test on a non-production server first.

Create a harmless scheduled command

For a Paper server:

1. Open Automations, then New automation.
2. Name it Player list check.
3. Select the in-game command action and enter list.
4. Set Cron expression to 0 * * * * and Timezone to UTC.
5. Leave retries disabled and choose to skip missed runs.
6. Choose Create automation.
7. Use Run now and inspect its run history and console output.

Expect a successful run and Paper's player-list response. Disable or delete this trial job when you no longer need it.

Choose a schedule

Use five-field cron syntax, with no seconds field:

The default timezone is UTC. Use an IANA timezone such as Europe/Berlin when the job follows local time. Consider daylight-saving transitions when scheduling maintenance.

Choose retries and catch-up deliberately

• No retries: one failed attempt remains failed.
• Fixed retry: repeats after the configured delay.
• Exponential retry: increases the delay, with an optional maximum.
• Skip missed runs: does not replay schedules missed while offline.
• Catch-up: runs up to the configured number of missed jobs on startup.

The default catch-up limit is three when catch-up is selected. Repeating a request can cause duplicate changes. Enable retries or catch-up only for an action that can safely run again.

Shell and web-request jobs

Shell jobs run as the agent user. Set the working directory explicitly for scripts. Raw-shell execution is off by default; enable it only when the script needs shell syntax.

Web requests originate from the agent environment. Set the method, headers, body, and timeout. The default timeout is 30 seconds, with a supported range of 1 to 300 seconds. Keep credentials out of shared job descriptions and screenshots.

Inspect failures

Read the run's status, reason, attempt, scheduled time, and failure. Confirm the agent had the required CLI, network access, and filesystem permissions.

For backups, use a tested backup script. A successful job exit does not prove that the backup can be restored.