Agent Browser Automation

•
3
•

Enables autonomous web navigation, interaction, and data extraction through a robust browser automation CLI designed for AI agents.

agent-browser is a specialized CLI tool that empowers Claude to interact with the web using Chrome/Chromium via the Chrome DevTools Protocol (CDP). It streamlines complex automation tasks by providing a unique element referencing system (@refs), allowing the agent to accurately click, fill, and navigate through dynamic web applications. With advanced features including encrypted authentication vaulting, session persistence, network request inspection, and visual diffing, this skill is ideal for developers needing to automate workflows on sites without APIs, perform end-to-end testing, or scrape data from authenticated sessions.

Características Principales

01Network & DOM Inspection: Monitors XHR/Fetch requests and captures detailed DOM snapshots.
02Live Session Streaming: Provides real-time WebSocket streaming to monitor browser actions as they happen.
03Advanced Auth Management: Includes an encrypted vault, session persistence, and profile reuse for secure logins.
043 GitHub stars
05Interactive Element Referencing: Uses unique @ref labels for reliable AI-driven UI interaction.
06Multimedia Capture: Generates annotated screenshots, PDFs, and full-page visual captures.

Casos de Uso

01Performing automated end-to-end testing and visual regression checks on web apps.
02Automating complex contribution workflows on platforms like Drupal.org or GitHub.
03Extracting structured data from interactive, JS-heavy websites that require authentication.