Agent Browser CLI Automation

•
2
•

Automates browser interactions, web scraping, form filling, and UI testing directly from the command line using interactive element references.

Agent Browser is a specialized browser automation CLI designed for AI agents to interact reliably with modern websites. By leveraging an interactive snapshot system that generates predictable element references (such as @e1 and @e2) alongside semantic locators, Claude can navigate pages, complete complex form submissions, bypass shell quoting hurdles for JavaScript execution, and preserve authentication state across sessions with built-in encryption. It is ideal for developers needing automated web research, end-to-end web application testing, dynamic data extraction, or visual verification across desktop Chrome and iOS Simulator environments.

Key Features

012 GitHub stars
02Session and authentication persistence with encrypted local storage and cookie handling
03Interactive snapshot ref system (@e1, @e2) and semantic locators for robust element selection
04Secure JavaScript evaluation using stdin heredoc and base64 encoding to prevent shell escaping issues
05Comprehensive page capture capabilities including full-page screenshots, PDFs, and video recording
06Multi-target execution supporting headless Chrome, existing remote CDP sessions, and iOS Simulators

Use Cases

01Mobile web compatibility testing and UI verification using Safari on the Xcode iOS Simulator
02End-to-end web testing and multi-step form automation with preserved authentication sessions
03Data extraction and content scraping from dynamic single-page applications and JavaScript-heavy sites