Monitors: Automate Recurring Agent Tasks

Monitors is the name you'll see in your dashboard for this feature. The instruction behind each one is called a standing order: a structured, versioned set of directions the agent follows every time it runs, not just a one-line cron command.

Four trigger types are supported:

• Cron: run at fixed times (e.g. every weekday at 9am), with an IANA timezone if you need one.
• Interval: run every N seconds, minimum 10.
• File watcher: run when a file or folder you've shared changes; filter by which kind of change (created, modified, moved, deleted) and by filename pattern.
• Immediate: fire once, right away (useful for kicking off a run from a webhook or another trigger).

How it works

Create a Monitor from the agent's settings, choose a trigger type, and point it at the task you want repeated. Each run is a full agent invocation, with the same reasoning, tools, and memory as any other conversation, so it adapts to what it finds rather than replaying a fixed script. The agent reports back through its normal channels.

An agent can also create its own Monitor mid-conversation, if a task turns out to need ongoing watching rather than a one-time answer.

Example

Trigger: cron
Schedule: 0 9 * * 1-5
Task: Summarize overnight support tickets and post to #support

Related

• Core Concepts: standing orders as the structured instructions behind every agent's behavior
• Agents: scheduled execution alongside autonomy levels and configuration history