Agent Browser CLI

•
9
•

Automates web interactions, visual testing, and form filling using a persistent, headed browser session within Claude Code.

The agent-browser skill provides a high-performance CLI for browser automation, serving as a powerful, token-efficient alternative to standard Playwright MCP tools. It enables Claude to navigate websites, capture screenshots, handle complex OAuth authentication flows, and perform visual UI verification with persistent session memory. By leveraging interactive element references (@refs) and compact snapshots, it allows for seamless interaction with web dashboards, forms, and dynamic content that traditional APIs or CLI tools cannot handle.

주요 기능

01Persistent session management to preserve authentication and browser state
02Comprehensive interaction suite including clicking, typing, scrolling, and file uploads
03Headed browser sessions for visual tracking and manual user authentication
049 GitHub stars
05Advanced visual verification with annotated and full-page screenshots
06Token-efficient DOM snapshots with interactive element tagging (@refs)

사용 사례

01Automating complex multi-step login flows and OAuth authentication
02End-to-end UI testing and visual verification of web applications
03Navigating and extracting data from interactive dashboards and JS-heavy sites