docs/usage/agent/scheduled-task.mdx
Scheduled tasks are jobs that run periodically in the cloud. In short, you can have an Agent run on your prompt on a schedule — for example, checking social media regularly and sending notifications. Instead of manually triggering the same workflow over and over, set it once and let it run automatically — daily, weekly, or hourly.
Find Scheduled Tasks in the left panel of the Agent conversation page, and click Add Scheduled Task to start creating a task.
Task name — Give the task a descriptive name so you can recognize it at a glance:
Task content — Enter the prompt or instructions the Agent should run each time the task fires. Be specific and complete — this exact prompt runs on every scheduled execution. For example:
Analyze today's top tech news and summarize:
1. Major product launches
2. Funding announcements
3. Industry trends
Format as a brief executive summary.
Frequency — Choose how often the task runs:
Time and timezone — Set the exact time and timezone so the task runs at the correct local time. Times use 24-hour format. For distributed teams, getting the timezone right matters.
Max executions — Optionally cap how many times the task runs in total. Ongoing tasks often need no limit; for time-boxed campaigns (e.g. 30 days), you might set 30 — the task disables itself after reaching the limit.
After you create a task, you can change its configuration at any time.
Daily morning report:
Weekly planning session:
Hourly monitoring:
End-of-month review:
Each scheduled run creates an entry in that Agent's conversation history, labeled with the task name and timestamp. You can review outputs, check for errors, and track past results.
Click a scheduled task to edit it — update the prompt, change frequency or time, or adjust the timezone. Changes apply from the next scheduled run onward.
If you temporarily don't need a scheduled task, turn off its enabled state. While off, it won't run automatically; the schedule and prompt stay saved. When you turn it back on, the task continues as configured.
If you no longer need a scheduled task, you can delete it. Deletion removes the schedule and prompt configuration; the system will not trigger further runs. Past conversation history is kept.
Write clear, self-contained prompts — The scheduled task prompt runs with no prior conversation context. Everything the Agent needs must be in the prompt:
Choose appropriate frequency — Match the schedule to how fast the information actually changes. Hourly checks for daily news add unnecessary load; weekly reports for real-time metrics miss important updates.
Use descriptive task names — Put purpose and timing in the name: "Weekly Competitor Analysis - Monday 9am" beats "Task 2".
Set max executions while experimenting — When testing a new scheduled task, use a max execution count of 5–10 so it doesn't run forever if the prompt needs tuning.
Timezone awareness — Always set the correct timezone. "09:00" is interpreted in the configured timezone, which may differ from your local clock. Wrong timezone is a common cause of unexpected run times.
Schedule a task to periodically check social content for given platforms or keywords. It can fetch recent activity, filter what matters, and summarize when there's something important — useful for brand monitoring, competitor tracking, or creator update alerts.
For work that needs regular review — analytics, project status, or content performance — a scheduled task can gather information on a cadence and produce structured takeaways so you keep sight of trends.
Set reminders for milestones, recurring checks, or follow-ups. LobeHub can generate reminder messages and notify you (for example by email) without you triggering the flow manually.
Task didn't run when expected — Check the timezone. Scheduled times are relative to the configured timezone, not necessarily "now" on your device. Also confirm the task is enabled.
Runs at surprising times — Double-check 24-hour time (e.g. 17:00 is 5:00 PM, not 5:00 AM).
Poor output quality — Scheduled prompts run without chat history. Rewrite the prompt so it is fully self-contained, with background, data sources, and format requirements spelled out.
Too many runs — While experimenting, set a Max executions cap. If a task has already run more than intended, delete it and create a new one with the right limits.
<Cards> <Card href={'/docs/usage/agent/web-search'} title={'Web Search'} /><Card href={'/docs/usage/getting-started/agent'} title={'Agent'} /> </Cards>