Triggers and runs

A trigger decides when an agent should run. A run is the execution record that shows what happened.

Trigger types

Custom agents can support manual runs and workflow-driven triggers. Depending on the workspace setup, triggers can also be scheduled or fired by changes in another product surface, such as a task-column change or inbound workflow event.

Use manual runs while testing. Use scheduled or workflow-driven triggers only after the agent has produced good results on a small sample.

Run history

Run history helps you evaluate reliability. A run can include status, summary, start time, completion time, duration, credits, tool calls, and failure details.

How to review a run

1. Check the status

   Confirm whether the run completed, failed, or is still in progress.

2. Read the summary

   Confirm the result matches the workflow goal.

3. Review tool usage

   Check whether the agent used the expected resources or tools.

4. Check output location

   Verify where the output landed, such as a record, task, draft, note, or handoff.

5. Inspect failures

   Missing context, unavailable integrations, timeouts, and unclear prompts are common causes.

When to pause a trigger

Pause or disable a trigger when:

• The connected resource is no longer correct.
• The prompt is being revised.
• The output requires too much manual correction.
• The agent is using unexpected tools.
• A downstream workflow or integration is unavailable.

Related guides