Testing your agent before it goes live

Run your agent on demand and watch exactly what it does — so you can catch issues and build confidence before it runs on its own.

Last updated July 14, 2026

Before you let an agent run on its own, it's worth watching it work once. Testing lets you kick off a run on demand and see exactly what your agent does, step by step, so you can catch anything off before it's live.

A quick test turns "I hope this works" into "I've seen it work."

Run a test

Start a test run

From the builder or the agent itself, trigger a run on demand instead of waiting for the schedule or trigger.

Watch it work

Follow the run as your agent moves through its steps and uses its tools. You can see what it's doing and why.

Check the result

Review what the agent produced — a draft, a report, a message — and confirm it's what you expected.

Adjust and re-test

If something's off, tell the builder what to change, then run another test until it's right.

Tip. A test run is context-aware and can be cancelled — if it starts heading the wrong way, stop it and adjust before trying again.

What to look for

  • Right tools, right accounts. Confirm the agent used the inbox, calendar, or CRM you intended.
  • Right output. Check the tone, format, and accuracy of what it produced.
  • Draft vs. send. Make sure anything sensitive — outbound email, messages, changes to records — behaves the way you want. If you meant "draft only," verify it didn't send.
  • Sensible handling of edge cases. If there was nothing to do, the agent should handle that gracefully rather than inventing work.
Heads up. A test is a real run — if your agent is set up to send messages or change data, it will. When in doubt, set it to draft or notify you first, then test.

Where test results show up

Every run, including tests, is recorded in Activity, and anything the agent produced lands in Deliveries. If a run fails, you'll find the details in Errors so you can see what went wrong and fix it.

Going live

Once a test run does what you want, you can trust the agent to run automatically on its schedule or trigger. You can always test again later after making changes.

Next steps

Was this page helpful?