Skip to main content

Run One Test

Start with one test when you are checking a new flow or investigating a failure.

Open the test and choose Run, pick Playwright or Assistant as the run method, confirm the environment and device profiles, use Change to check the tests in Choose tests, expand Run options for concurrency, and start the run, which opens with its own link. The same single test is then dispatched from the terminal with --wait, which prints the summary and exit code when the result arrives.

Start a run​

  1. In Tests, open the test and choose Run. For a draft, choose Assistant run, or generate its Playwright code first.
  2. In New run, choose a Run method:
    • Playwright executes the saved test code. The test must have runnable code.
    • Assistant follows the saved Gherkin or Markdown instructions against the app. It uses Assistant credits and does not require generated code.
  3. Confirm the Environment and Device profiles.
  4. For an Assistant run, choose Model, or leave Project default selected.
  5. Review the Tests summary. Change opens Choose tests; search or expand folders, check the tests to include, and choose Done. An empty selection shows Choose tests directly.
  6. Choose Start run or Start Assistant run.

You can also open Runs, select New run, and choose the test there. The same composer lets you review the selection before anything starts.

Choosing tests in the run composer

Options and results​

For Playwright, expand Run options to change concurrency or retry failed scenarios. Auto uses the project default, Sequential runs one scenario at a time, and Custom sets a cap for this run. Assistant runs follow scenarios sequentially.

Make sure the selected environment is reachable and required variables and secrets exist. Private apps need an online tunnel device.

The run opens with its own link. Results load separately from the run summary; a loading message is not an empty or failed result. Open a scenario for its error, progress, and available evidence. See Understand live run status.

From the command line​

With the CLI authenticated and a project selected, run a feature by name:

testvibe run "Sign in" --wait

--wait waits for the result within the CLI's wait budget and prints a summary. A failed result returns a nonzero exit code. Without --wait, the command returns after dispatch. A wait timeout does not cancel the server-side run; use testvibe runs show <runId> to check it.

See Review results for the next step.