Your first test
Write a three-step test in plain English, run it, and read the result.
Write the test
Open your project, pick a suite, and add a test. A test is a name plus a set of plain-English steps. A good first test checks one user journey end to end:
1. Go to the home page and click "Log in".
2. Sign in with the test account credentials.
3. Check that the dashboard greets the user by name.Keep each step to one action or one check. The agent reads the page like a person, so refer to things by their visible text ("the Log in button"), not by technical detail.
Run it
Run the test from the app. Mockingbird queues the run, opens a fresh, isolated browser session, and works through your steps against the project's base URL.
Read the result
When the run finishes you get:
- A verdict: passed or failed, with the agent's reasoning for the judgement.
- A step timeline: every action the agent took, with a screenshot at each step.
- A session recording of the browser session, so you can replay exactly what happened.
If a step could not be completed, the timeline shows where the agent got stuck and what it tried. Failures read like a bug report from a careful tester rather than a stack trace.
Runs have a per-step time budget and a per-plan cap on total steps, so a stuck test ends with a clear timeout message instead of hanging. See Plans and limits.