Agent Browser Automation
Automates headless web interactions, form filling, and data extraction using a reference-based CLI system optimized for AI agents.
Agent Browser is a specialized automation skill that leverages Vercel's agent-browser CLI to provide a high-fidelity interface for web interactions within Claude Code. By utilizing ref-based selection from accessibility snapshots (e.g., @e1, @e2), it enables Claude to navigate websites, fill complex forms, capture screenshots, and scrape data with extreme precision. This skill is particularly useful for workflows requiring reliable element targeting without the overhead of complex selector strings, offering a robust alternative to standard browser automation tools for quick tasks and deep web research.
Key Features
01Detailed JSON snapshots of interactive web elements
02Integrated screenshot and PDF generation for visual verification
03Ref-based element selection from accessibility snapshots for high precision
04Full browser control including navigation, clicks, and form filling
053 GitHub stars
06Support for parallel browsing sessions and headed debug modes
Use Cases
01Automated web scraping and data extraction from dynamic sites
02Generating visual documentation or PDFs from live web pages
03End-to-end testing of user flows like login and checkout