Skip to main content

​
Overview

Scheduled Tasks let you run any agent automatically on a recurring schedule. Create a task in Control Hub, pick an agent, write a prompt, set a schedule, and let it run on its own. Every execution starts a new conversation with the agent, just like sending a message manually.

​
How It Works

Create the task

You create a scheduled task with an agent, a prompt (the message to send), and a schedule.

Aster checks what's due

Aster runs a job every minute that checks which tasks are due.

The prompt is sent

When a task is due, it sends the prompt to the agent as a new conversation.

The agent runs

The agent runs with all its tools, knowledge bases, and capabilities — just like an interactive chat.

The conversation appears

The resulting conversation appears in the dashboard with source scheduled-task.

​
Creating a Scheduled Task

​
From Control Hub

Open Scheduled Tasks

Navigate to Control Hub → Scheduled Tasks, then click “Create Task”.

Fill in the form

  • Task Name: A descriptive name (e.g., “Daily VP Candidate Digest”)
  • Agent: Select which agent should run
  • Description: Optional notes about what this task does
  • Schedule: Use a preset or write a custom cron expression
  • Prompt: The message the agent will receive when the task runs

Create the task

Click Create Task. The task starts enabled by default and will run on its next scheduled time.

​
From a Conversation (Manage Scheduled Tasks tool)

Agents with the manage_scheduled_tasks tool enabled can create scheduled tasks during conversations. Just ask:
  • “Check stock prices every morning at 9 AM”
  • “Send me a weekly summary of new candidates every Monday”
  • “Run a daily report and email it to the team”
The agent will create the scheduled task for you. Enable the manage_scheduled_tasks tool in your agent’s configuration under Tools.

​
Schedule Options

​
Presets

Click any preset button in the task form for common schedules:

​
Custom Cron Expressions

For more control, enter a standard 5-part cron expression:
Examples: The form shows a live preview of when the task will run as you type.
Schedule times are in UTC. If you create tasks via the manage_scheduled_tasks agent tool with a timezone provided, the time is automatically converted to UTC.

​
Managing Tasks

Enable / Disable

Toggle any task on or off with the switch in the task list. Disabled tasks won’t run but keep their configuration and history.

Run Now

Trigger any enabled task immediately from the actions menu (three dots → Run Now). This creates a new execution just like a scheduled run.

Edit

Update the name, agent, prompt, or schedule of any task. Changes take effect on the next scheduled run.

Delete

Remove a task permanently. Execution history is preserved for reference.

Admin View

Organization admins can toggle Admin View to see all scheduled tasks across the organization, not just their own.

​
Execution History

Click View History in the task’s actions menu to see past executions:
  • Status: Whether each run completed successfully or failed
  • Timestamps: When each execution started
  • Conversation Links: Click through to the conversation created by each run
  • Stats: Total runs, completed, failed, and success rate

​
Use Cases

Create a task that runs daily and emails a summary to your team.Task prompt example:
Have an agent query your database and generate a report on a schedule.Task prompt example:
Schedule an agent to check for issues and notify your team.Task prompt example:

​
Combining with Other Features

Scheduled tasks work with all agent capabilities:

​
Best Practices

  1. Write clear, specific prompts. The agent receives only the prompt text — it doesn’t have context from previous runs. Include all the instructions it needs in the prompt itself.
  2. Use the right schedule frequency. Don’t run tasks more often than needed. A daily digest doesn’t need to run every 15 minutes.
  3. Test with Run Now first. Before relying on the schedule, use Run Now to verify the agent produces the output you expect.
  4. Check execution history. Review past runs periodically to catch failures or unexpected behavior.
  5. Include error handling in prompts. Tell the agent what to do when there’s nothing to report (e.g., “If no new data, send a brief confirmation email instead”).