Launches
A run of a test suite and its results.
A launch is a single run of a test suite. It
runs every test case in the suite — each in its own cloud browser, in parallel —
and records a per-test verdict, the agent's step-by-step logs, and screenshots.
Each launch belongs to one suite and gets a sequential number within it, so runs
are named Suite name #3. Running the same suite again never overwrites an
earlier launch — it creates a new one with the next number, and past launches are
kept as history.
Launch statuses
A launch starts in progress and settles into the final status once every test case has finished.
| Status | Meaning |
|---|---|
| In progress | The run is still executing. Test cases are running or queued. |
| Passed | Every test case met its criteria. |
| Failed | At least one test case did not meet its criteria — the product behaved incorrectly. |
| Error | A test case couldn't complete because of an agent or infrastructure problem (for example the sandbox failed to start), and none failed on product criteria. |
The distinction between failed and error matters when you read results. Failed is a verdict about your product: the agent reached a conclusion and the app didn't do what the test case described. Error is a problem with the run itself, so there's no product verdict to trust.
Reading a launch
Open Launches in the project sidebar to see every run, newest first, with its ID, name, start time, and status. Filter by suite, status, or trigger to narrow the list. Selecting a launch opens its detail view, which has three parts:
- Results — passed, failed (including errors), and still-running counts at a glance.
- Launch details — the run's metadata: status, ID, the suite it ran, what triggered it, start and end times, and total duration.
- Tests — one collapsible row per test case, ordered so anything still running or failing sits at the top. Each row shows the case's ID, name, screen size, and status. Expand a row to see the URL it opened (with variables filled in), the description, timing, and the full log.
The test log
Inside each test, the log is the agent's narration of the run — the same stream you watch live in preview mode. Reading it top to bottom shows exactly how the agent reached its verdict.
| Log entry | What it is |
|---|---|
| Info | A neutral note, such as navigating to a URL. |
| Reasoning | The agent's thinking about what to do next. |
| Action | A concrete step it took (clicking, typing, etc.). Most actions include a screenshot of the page at that moment. |
| Passed / Failed | The final verdict for the test case and the reason behind it. |
| Error | A run error that stopped the test before it could reach a verdict. |
Click any screenshot to open it full size. While a launch is in progress, its detail view streams new log entries and status changes live, so you can watch a run unfold without refreshing.
How is this guide?