Scheduled Tasks
You can ask the Main Agent to run a task on a recurring schedule. Once scheduled, the agent re-runs the task at the configured time, executes the plan in Auto mode, and posts the results back to the task detail.
When to Schedule
| Use Case | Example Schedule |
|---|---|
| Daily brand brief | Weekdays 9:00 AM |
| Weekly social media kit | Mondays 6:00 AM |
| Monthly campaign render | First of the month |
| Recurring report from a Monday board | Every Friday 5:00 PM |
| Drip-style email broadcast | Every 3 days |
Creating a Schedule
- Open the task you want to schedule (or create a new one).
- Click Schedule in the task detail header.
- Choose a cadence (preset or custom cron expression).
- Pick a timezone.
- Save. The task is added to the Scheduled section of the task list with the next run time.
Managing a Schedule
Open the same Schedule panel on a scheduled task to manage it. Every field is fully editable after creation:
| Action | What it does |
|---|---|
| Edit cadence | Switch presets or write a custom cron expression — the next-run time recalculates immediately |
| Edit timezone | Change the timezone the cron expression is evaluated in |
| Pause | Stops future runs without losing the schedule. Resume to pick up at the next tick |
| Resume | Re-enables a paused schedule |
| Delete | Removes the schedule entirely. Past runs stay in the task timeline |
| Next run | Always visible on the task card so you know when the agent will fire next |
How Execution Works
Schedules run as cron jobs in the platform. Each tick:
- Checks for due schedules in order of
next_run_at. - Acquires a lock per task so the schedule never double-fires.
- Spawns a fresh agent run using the saved task brief.
- Executes in Auto mode and writes results back to the task.
The processor batches up to 100 due tasks per tick to keep latency tight.
info
Schedules created in the agent UI run in the platform's cron infrastructure. If the agent service is briefly unavailable, schedules catch up at the next tick — they don't silently skip.
Combining with Connectors
Schedules pair well with integrations:
- Daily Monday digest — every morning, query the Monday board, summarize blockers, and post an update.
- Weekly Jira release notes — every Friday, pull resolved tickets and draft release notes.
- Recurring Drive sync — every hour, scan a Drive folder for new briefs and spawn child tasks.
Limits
- Concurrent schedules per workspace depends on your plan.
- Minimum interval is 5 minutes for Enterprise; lower-tier plans cap at hourly.
- Failed runs are retried once, then marked failed in the task feed for review.
Related Pages
- Integrations — trigger sources you can pair with schedules
- Skills — recipes that schedules will run
- Models — pick the right model tier for scheduled cost-routing