Agent Browser Automation

•
10
•

Automates headless web browser interactions using ref-based element selection and accessibility snapshots for precise AI-driven navigation.

The agent-browser skill empowers Claude to perform complex web tasks via Vercel's specialized CLI. By generating accessibility snapshots that map interactive elements to simple references (like @e1 or @e2), it allows the AI to navigate pages, fill forms, and scrape data with high reliability. This skill is essential for workflows requiring real-time web access, multi-step automation, or interaction with dynamic JavaScript-heavy sites that traditional scrapers cannot handle.

主要功能

01Ref-based element selection using accessibility snapshots
02Full browser control including navigation, clicks, and typing
03Visual verification through screenshots and PDF exports
04Parallel session management for multi-site workflows
05Semantic locators for role-based and text-based interactions
0610 GitHub stars

使用场景

01Automating multi-step login and data entry workflows
02Scraping dynamic content from SPAs and JavaScript-heavy websites
03Performing end-to-end web testing and visual regression checks