Webwright Playwright Automation
Automates multi-step web browsing tasks using local Playwright execution, visual screenshot verification, and reproducible Python scripts.
Webwright adapts the code-as-action automation paradigm directly into Claude Code, executing local headless Playwright browser workflows step-by-step. It breaks user requests into verifiable critical checkpoints, navigates dynamic websites using robust DOM and ARIA snapshots, records execution logs and screenshots, and leverages Claude's native vision to inspect results against strict criteria. By avoiding external vision APIs or fragile one-shot scrapers, Webwright reliably delivers both immediate extraction results and maintainable, parameterized Python CLI scripts.
Key Features
011 GitHub stars
02Strict constraint enforcement for dynamic UI filters, sorts, modals, and pagination
03Self-verification loop using Claude native vision on step screenshots
04Dual operation modes supporting one-shot execution and parameterized CLI tool generation
05Step-by-step local Playwright automation in headless Firefox
06Structured workspace contracts tracking run logs, scripts, and visual evidence
Use Cases
01Automating multi-step web tasks like complex form submissions, search filtering, and portal navigation
02Generating reusable, maintainable Playwright CLI automation scripts for repetitive developer workflows
03Extracting structured data from anti-bot or dynamic web applications with visual proof of accuracy