Testboxes

Choose the test target

Open Engineering → Testboxes. UI and smoke tests target an application URL; integration tests can also read a repository; unit tests target a repository and its test command. Choose a deployment, model, and optional GitHub integration.

UI, smoke, and integration tests use a Playwright image. Unit tests use the devbox default image and the repository's own test runner. The repository is checked out under /workspace/repo.

Generate and run tests

Write cases with IDs, names, descriptions, and plain-language steps in the Plan tab. Generate tests starts the bound agent; Open chat lets you refine its work. For browser tests, it explores the application and maintains Playwright specs alongside testbox.pilot.yaml.

Run the suite on demand or set a phrase, interval, or cron schedule. A testbox with a run already in progress skips its scheduled tick. Results retain per-test output and, where applicable, screenshots, traces, and an HTML report.

Investigate failures

Use Ask the agent to fix on a failed run. The agent receives the failure messages, reproduces browser failures where applicable, and distinguishes product bugs from brittle tests before making changes and rerunning the suite. Inspect the resulting tests and artifacts.

Login passwords are encrypted and supplied to runs as TESTBOX_USERNAME and TESTBOX_PASSWORD. API routes live under /api/v1/dashboard/testboxes; reads require devboxes:read and mutations require devboxes:write.