Scheduled Runs

Use scheduled runs to execute saved scripts on a fixed cadence without manual dispatch.

Path: App -> Schedules

Create a schedule

1. Open /app/schedules.
2. Choose script and target agent.
3. Enter a cron expression and timezone.
4. Set guardrails:
   • max runtime minutes
   • max retries
   • retry delay seconds
   • concurrency limit
   • idempotency window minutes
5. Save.

Default runtime is 12 hours (720 minutes). Schedules can be configured up to 22 hours (1320 minutes) when a workflow legitimately needs a longer timeout.

Configure variables and vault keys

• Put non-secret inputs in variables (JSON object).
• Put secret inputs in vaultKeys so the scheduler injects decrypted values at runtime.
• Missing vault keys block dispatch for that run.

Trigger behavior

The scheduler evaluates due rows and applies these controls before dispatch:

• organization billing gate
• concurrency limit per schedule
• idempotency window
• agent readiness (including auto-start when stopped/offline)

If checks pass, Mimic dispatches a job and writes schedule metadata (lastRunAt, lastRunStatus, nextRunAt).

Manual and dry-run execution

Use POST /api/schedules/execute for scoped manual execution.

Example:

{
  "dryRun": false,
  "force": true,
  "scheduleId": "sched_abc123",
  "limit": 1
}

Rules:

• force: true requires scheduleId.
• dryRun: true reports what would trigger without dispatching jobs.
• Manual runs now return detailed triggeredSchedules, skipped, and errors data so the UI can show lock contention, agent readiness, or vault problems directly.

Cron worker endpoint

Use GET or POST /api/cron/scheduled-runs with x-cron-key (or bearer token) set to CRON_SECRET.

Recommended:

• run every minute
• include dryRun=true in non-production smoke checks
• alert on repeated executor lock errors

Operating model

• Keep deterministic selectors and explicit assertions to preserve RPA-first reliability.
• Use retries for transient portal failures only.
• Use stale-job watchdog endpoints alongside schedules to keep queues recoverable.