Schedule Recurring Tasks and Automations

Before You Start

Use scheduled jobs for work that is already repeatable: daily inbox triage, weekly account research, monthly reporting, renewal reminders, or checks that should run when a connected app event happens.

Before scheduling, make sure the Super Agent or target agent can complete the task manually. Connect the required apps, confirm the agent has permission to read or change the right records, and decide which actions should require human approval.

Have the expected output, destination, owner, and cadence ready. For example, know whether the result should be a dashboard summary, Slack draft, Linear issue, Gmail draft, or internal note.

Ask in Plain Language

Open your Super Agent and describe the automation the same way you would delegate it to a teammate. Include the goal, source system, timing, output format, delivery destination, and approval rule.

For example: "Every weekday at 8:30 AM Eastern, review new Gmail support messages, group them by urgency, draft replies for billing questions, and post a summary to the support Slack channel. Ask before sending any customer reply."

If you already have a sub-agent for the job, name it in the request. If not, the Super Agent can help create one first or schedule the workflow directly when the responsibility is narrow enough.

Use a Template or Dry Run First

The Super Agent may propose a reusable template before it creates the schedule. Review the template for stable inputs, clear success criteria, tool access, and any steps that must stay approval-only.

Run the task manually or ask for a dry run before the schedule goes live. A dry run should show what the automation would read, what it would produce, and which actions it would ask you to approve.

Do not schedule broad or ambiguous prompts. If the dry run needs repeated corrections, update the template or agent instructions and test again before enabling the recurring job.

Choose a Schedule or Trigger

Use a time-based schedule when the work should run on a predictable cadence such as hourly, daily, weekly, monthly, or a custom cron expression. Cron schedules are useful when you need a specific pattern, such as every weekday morning or the first business day of the month.

Use an event trigger when the work should start after something happens in a connected app, such as a new message, updated issue, form submission, paid invoice, or changed customer record. Event-triggered jobs depend on the connected app and permissions available in your workspace.

Keep the first schedule conservative. Start with the least frequent cadence that still meets the business need, then increase frequency after the first successful runs.

Confirm Timezone and Ownership

Always confirm the timezone before enabling a time-based job. Use the business timezone for team reports and customer commitments, not your browser timezone, unless the task is personal to you.

Name an owner for the automation. The owner should know what the job does, where results appear, what approvals are expected, and when to pause or edit the job.

If teammates work across regions, include the timezone in the job name or description so everyone can read the schedule without guessing.

Understand Credits and Limits

Scheduled jobs consume credits when they run, just like manual agent work. Tool-heavy workflows, large files, frequent schedules, and long-running analysis can use more credits than short checks.

Before enabling a high-frequency job, review your plan and available credits. Pause or lower the cadence for automations that no longer create enough value for their usage.

If your workspace uses a connected ChatGPT account for eligible Super Agent runs, confirm that behavior with a small scheduled test and review the trace, the step-by-step run log, if usage looks different from expected.

Manage Scheduled Jobs

Open Scheduled Jobs to review active and paused automations. From there, you can inspect the job, change the prompt or template, adjust the cadence, update the timezone, pause the job, resume it, or delete it when the workflow is no longer needed.

Pause a job before changing connected-app permissions, rotating credentials, or changing a process the automation depends on. Resume it after a manual test confirms the updated setup still works.

Use clear names such as "Weekday support triage - 8:30 AM ET" or "Monthly renewal risk report - first Monday" so teammates can understand the job list quickly.

Verify the First Runs

After enabling the job, check execution history after the first scheduled run. Confirm it ran at the expected time, used the intended agent, read the right sources, respected approval rules, and produced an output your team can act on.

Open the trace for the first few runs. Look for missing inputs, failed tool calls, skipped approvals, timezone mismatches, or output that needs clearer formatting.

Once the job is stable, review it periodically. Scheduled work can drift when business rules, connected apps, permissions, or team ownership changes.

Troubleshooting

If the job does not run, confirm it is enabled, the schedule or trigger is valid, the timezone is correct, and the owner still has access to the workspace.

If a cron schedule fires at the wrong time, check the timezone and cron expression together. Update both in the job settings, then wait for the next scheduled window or run a manual test.

If an event trigger does not fire, reconnect the app, confirm the app event is supported, and verify the connected account can see the source record.

If the job runs but cannot use a tool, reconnect the integration or enable the required tool on the agent. Then rerun the template manually before relying on the next scheduled execution.

If credits are used faster than expected, lower the cadence, narrow the prompt, reduce unnecessary tool calls, or pause the job while you review plan limits.

If the output is inconsistent, update the template with a stricter format, add examples, or inspect the trace to see where the agent needed more context.

Related Guides

Read "Meet Your Super Agent Command Center" if you want more context on using the Super Agent as your command center before scheduling work.

Read "Connect Apps and Tools to Your Agents" before scheduling work that depends on Gmail, Slack, Linear, Stripe, or another integration.

Read "Create Your First Agent With Your Super Agent" when the automation needs a dedicated role instead of a one-off Super Agent request.

Read "Create and Reuse Prompt Templates" when the request should be saved and reused before it becomes a scheduled job.

Read "Read an Agent Trace" to debug failed runs, missing tool access, and behavior that changes between executions.

Read "Manage Credits and Plan Limits" when you need to understand credits, usage, and plan limits for recurring work.