browser_generate_playwright_test turns the clicking-around into code, so a bug reproduced once becomes a test that reproduces it every time afterwards, rather than a paragraph describing what someone did. The console messages and network requests come from that same session, which means the evidence and the repro are the same run instead of two attempts that may not match.
A browser automation server built on sessions: 29 tools covering launch, navigation, interaction by CSS selector, dialogs and tabs, screenshots and PDF, network and console capture, and Playwright test generation.
- Sessions launched headed or headless, with viewport, timeout, screenshot quality and compression chosen at launch
- Interaction by CSS selector: click, type, hover, drag between two selectors, key press, dropdown select and file upload
- Waiting on a selector, plus three readings of the page — accessibility tree snapshot, HTML content and visible text
- Navigation and tabs: back, forward, refresh, resize, and new, list, select and close for tabs
- Dialog handling for alert, confirm and prompt, with the prompt text supplied
- The page's own record of the session: every network request and every console message
- Playwright test code generated from the actions recorded in that session
Npx on your PATH, and the sessionId returned by browser_launch, which every subsequent call takes. No account, no key.
One command — npx @snowfort/circuit-web@latest