Skip to main content
Agent Manager supports scheduling recurring or one-time agent runs without any external job orchestration. Scheduled runs execute as background jobs and produce the same RunResponse as manually triggered runs, so they work with all the same monitoring and audit tooling.

Common use cases

  • Daily report generation (summarize overnight activity, generate executive briefings)
  • Periodic data analysis (run analytics agents on a nightly or weekly cadence)
  • Automated monitoring alerts (poll data sources and notify on anomalies)
  • Batch knowledge base updates (re-ingest and refresh vector content on a schedule)

Creating a schedule

1

Define the schedule payload

Build a ScheduleDTO with the agent to run, the message to send it, and the recurrence expression:
2

Submit the schedule

The API returns 201 Created with the new schedule including its assigned ID.

Cron expression reference

Schedules use standard five-field cron syntax:

Managing schedules

List all schedules (paginated):
Get a specific schedule:
Update a schedule (change message, recurrence, or enabled state):
Delete a schedule:
Returns 204 No Content on success.

Triggering a schedule manually

To execute a schedule immediately without waiting for its next cron time:
Manual triggers enqueue the same background job as an automatic execution.

Viewing run history

To see the execution history for a specific schedule:
Each entry in the list corresponds to one execution and includes its status, start time, and outcome. You can also monitor background jobs produced by scheduled runs via the Jobs dashboard.
Scheduled runs execute as background jobs. If a scheduled run fails, the job status will show FAILED and the failure reason will be captured in the run record. The schedule itself continues to execute on its defined cadence regardless of individual run outcomes.

Batch jobs

Agent Manager also supports spot batch jobs for one-off bulk operations. View all batch jobs:
Each batch job entry includes its status, progress percentage, estimated cost, and the compute configuration it ran on.
For time-sensitive schedules, use the manual trigger endpoint during testing to verify the agent produces the expected output before relying on the cron schedule in production.