What problem does it solve? Automating web UIs—filling forms, clicking through wizards, dismissing modals, and verifying page state—requires a reliable way to identify elements and handle sessions, auth, and dynamic DOM changes without brittle selectors. ## Core Features & Use Cases - Snapshot-driven interaction: Always capture an accessibility snapshot to discover element refs before clicking, typing, or filling forms, avoiding stale or guessed selectors. - Session and auth handling: Choose between the VS Code integrated browser, Playwright MCP with --extension for reusing a logged-in Chrome session, or persistent profiles via --user-data-dir. - Opt-in capability groups: Enable network mocking, storage manipulation, coordinate-based mouse control, PDF export, and tracing/video via --caps flags. - Use Case: Automate a multi-step checkout wizard on a staging site behind HTTP basic auth: navigate with embedded credentials, snapshot each step, fill form fields by ref, and verify the confirmation page. ## Quick Start Ask the agent to open the target website, fill in the login form, and click through the checkout flow using Playwright browser automation.