What problem does it solve? Cloning, QA-ing, or reproducing a running web app requires the exact state users actually see — fixture data, conversation text, status pills, queue order, keyboard shortcuts — which curl and static HTML scraping miss when content is client-routed or rendered only after JavaScript runs. ## Core Features & Use Cases - Live Browser Capture: Drive a real browser session with goto_url, js evaluation, screenshots, and CDP calls to read rendered DOM state. - Reliable SPA Navigation Pattern: Click table rows by index and poll location.pathname instead of using fixed sleeps, avoiding stale reads after React router.push navigation. - State Pollution Prevention: Clear localStorage and cookies between captures so your own exploration does not alter what the site shows on later loads. - Use Case: When reverse-engineering a Next.js review queue, capture each lead's detail page by clicking rows in order, polling for navigation, and saving main.outerHTML per item to the workspace for a 1:1 clone. ## Quick Start Use browser_exec to open the live site, click through each table row while polling for navigation, and save the rendered HTML of every detail page to the workspace.
