What problem does it solve? Manually documenting real browser workflows for RPA migration is slow and error-prone: engineers must capture clicks, form inputs, network requests, and page states across complex SPAs and strict-CSP sites, then translate them into automation steps for tools like UiPath, Power Automate, Selenium, or Playwright. ## Core Features & Use Cases - Browser Flow Recording: Launch a real Chrome session via Playwright, capture interactions, Fetch/XHR requests with full headers and bodies, navigation, and errors into per-session JSONL files. - AI Page Perception and Control: Observe pages (text, ARIA snapshot, interactive elements, screenshots) and execute click/fill/type/press/select/check actions through a unified control API with ambiguity detection. - Trace Conversion and Refinement: Convert session.jsonl into a lossless markdown trace, then refine it into an RPA draft with business step naming, selector sets, and noise annotation. - Use Case: Record an Oracle registration flow once, generate trace.md with all selectors and network assertions, then hand the refined rpa-draft.md to engineers building the UiPath or Playwright automation. ## Quick Start Ask the AI to start recording a browser session for a target URL, perform the workflow, then stop and convert the session into an RPA trace document.
