playwright

Playwright

End-to-end browser testing (Chromium, Firefox, WebKit) plus an AI authoring loop — the Test Agents. Everything runs through npx; do not assume a global install and do not npm i -g playwright. If a project already depends on @playwright/test, prefer its local version (npx playwright ... resolves it).

Docs: https://playwright.dev — Test Agents: https://playwright.dev/docs/test-agents

First contact

Always discover the live command surface rather than guessing — it's version-specific:

npx playwright --help              # all commands for the resolved version
npx playwright <command> --help    # options for one command
npx playwright --version           # confirm version (Agents need >= 1.56)

If browsers aren't installed yet, the first run prompts for them:

npx playwright install             # download Chromium/Firefox/WebKit
npx playwright install chromium    # just one engine (faster)
npx playwright install-deps        # Linux/CI OS deps (asks for sudo)

Core CLI

npx playwright test                      # run all tests
npx playwright test --grep "login"       # filter by title
npx playwright test path/to/spec.ts      # one file
npx playwright test --project=chromium   # one browser
npx playwright test --headed             # watch it run
npx playwright test --ui                 # interactive UI mode (debug, time-travel)
npx playwright test --debug              # step through with the inspector

npx playwright codegen localhost:3000    # record clicks → generated test code
npx playwright show-report               # open the last HTML report
npx playwright show-trace trace.zip      # open a saved trace (DOM + network + screenshots)
npx playwright screenshot <url> out.png  # one-off screenshot
npx playwright open <url>                # open a page in the bundled browser

trace: 'on-first-retry' and screenshot: 'only-on-failure' in playwright.config.ts are the high-value defaults — they make show-trace and the HTML report useful for diagnosing failures without slowing the happy path.