Agent Browser Core
Automates complex web browser tasks using an efficient snapshot-and-ref system to minimize token usage for AI agents.
The Agent Browser Core skill provides Claude with a high-performance interface for Chrome and Chromium automation via the Chrome DevTools Protocol. By utilizing accessibility-tree snapshots and compact element references (@eN), it allows agents to interact with web pages—performing actions like clicking, typing, and data extraction—using significantly fewer tokens than traditional HTML parsing. This skill is essential for developers building AI agents that need to navigate multi-tab sessions, handle secure authentication, capture screenshots, or perform automated end-to-end testing with high reliability and speed.
Key Features
01Token-efficient accessibility-tree snapshots and element referencing
02Support for multi-tab management and parallel browser sessions
03Built-in authentication vault and session persistence capabilities
04Advanced network mocking, request routing, and HAR recording
05High-fidelity screenshot and video recording with ref-based annotation
0635,846 GitHub stars
Use Cases
01Automating complex end-to-end testing for web applications
02Scraping structured data from dynamic, JavaScript-heavy websites
03Streamlining web-based workflows like account creation and form submission