Key Concepts
Workspace is the shared home for projects, members, billing, credits, and storage. Project is one app's tests, environments, settings, and results. Choose the correct workspace and project before you start.
Tests and instructions
A group is a folder in Tests and can contain nested groups. A test feature contains the authored instructions and any associated executable code.
Write instructions as Gherkin (Given, When, Then) or Markdown. A Gherkin scenario is one behavior to verify; a Background contains setup repeated before every scenario in that feature. Group membership does not establish execution order or shared browser state.
Variables and secrets supply values at runtime. The editor shows references as chips; the saved text uses {{var:NAME}} or {{secret:NAME}}. Use secrets for sensitive values and {{unique}} in values that need a fresh per-run stamp.
Assistant and generation
The Assistant is a project conversation for questions, investigation, suggestions, and proposed changes. A model is the selected provider model; project defaults and task overrides live in Settings → Assistant Model.
Generation turns saved instructions into Playwright code and verifies it against the app. Its conversation shows messages, questions, and verification outcomes. Regenerate starts a fresh attempt and can replace manual edits to generated code.
A proposal is a change ready for you to review. Preparing it does not mean it has been applied. A suggested test may be a draft or have verified code; check the state shown before accepting or running it.
Runs and environments
An environment is the app URL and associated configuration you are testing. A tunnel device lets cloud sessions reach an app inside a private network.
A Playwright run executes saved code. An Assistant run follows the saved instructions directly and has a model override in the run composer. Device profiles choose the browser size/configuration used for that run.
A run can cover one test feature or a selection of features. A feature may contain several scenarios; multiple profiles add more result rows. Re-run opens a pre-filled composer and creates a new run only when you start it. Cancel run stops active execution.
An automation starts work on a schedule or when a run finishes. See Automations.
Results and source code
Passed, Failed, and Skipped describe scenario outcomes; Pending and Running are live states. Skipped means a scenario did not complete; it may not have started or may have been cancelled after starting. Check the run-level context. A failed run can include passes or fail before any test executes. A result-fetch error is different from a test failure.
Evidence can include screenshots, videos, Playwright traces, browser console/network records, and uploaded artifacts. Availability depends on the run and capture settings.
Application shows connected application source context. Settings → Repositories manages source connections and optional test-file sync. Reading app source and mirroring test files are separate features.
Start with the Quickstart, Tests overview, or Review results.