Skip to main content

​
Test mode

Use the Test tab in the agent builder to run controlled tests before making an agent live.

​
Test flow

  1. Select the target user
  2. Select a recent event example for the configured trigger event
  3. Review or edit the event payload JSON
  4. Click Run test

Agent test flow

Test mode runs the agent with your selected payload and writes a test run you can inspect in the builder.

​
Event payload behavior

When recent events are available for the trigger event, Flywheel preloads a real event example. When no recent events are available, Flywheel falls back to schema-based payload generation so you can still run tests.
If your trigger event is not configured yet, set it on the New Event trigger in Editor first.

​
Run history and details

Use the Runs tab to inspect run behavior:
  • Run History: list of recent runs for the selected node context
  • Logs: chronological execution logs
  • Debug: step timeline and step result data
This is the fastest way to validate prompts, tool selection behavior, and output quality.

​
Draft vs published behavior

Agent runs should be validated with your current draft and then published when ready:
  • Publish changes applies current draft graph/config to the live version
  • Discard changes restores the last published graph/config
  • Live toggle controls whether real trigger events execute the agent
Before enabling Live:
  1. Test with at least one representative user
  2. Test with at least one edge-case payload
  3. Verify required tools return expected outputs
  4. Verify messages/emails match tone and policy
  5. Confirm logs and debug timeline have no unexpected failures