Agent-Browser Core

0•

Automates web browsing and interaction for AI agents using a fast, token-efficient accessibility-tree snapshot workflow.

The Agent-Browser Core skill provides a robust CLI-based interface for AI agents to control Chrome and Chromium browsers directly via the Chrome DevTools Protocol (CDP). By utilizing accessibility-tree snapshots and compact element references, it allows agents to navigate pages, interact with UI elements, and extract data while significantly reducing token consumption compared to raw HTML parsing. It serves as the foundational toolkit for web automation tasks including form filling, automated testing, data extraction, and authenticated session management across multiple tabs or parallel browser instances.

主な機能

01Visual debugging tools including annotated screenshots and video recording
02Token-efficient accessibility-tree snapshots using @eN references
03Comprehensive session management with auth vaults and state persistence
04Direct browser interaction (click, fill, type, select) without Playwright/Puppeteer dependencies
05Advanced waiting logic including network idle, element visibility, and URL patterns
060 GitHub stars

ユースケース

01Executing end-to-end UI tests and troubleshooting web application failures
02Automated web scraping and data extraction from complex single-page applications
03Automating multi-step web workflows like account setup, form submission, and reporting