Agent Browser

0•

Automates browser interactions and UI testing using AI-friendly accessibility snapshots and deterministic element references.

Agent Browser is a specialized headless browser CLI designed specifically for AI agents, replacing fragile CSS selectors with a robust 'snapshot-then-act' workflow. By generating accessibility snapshots with unique refs (e.g., @e2), it allows Claude to identify and interact with UI elements with high precision, making it ideal for troubleshooting visual bugs, performing end-to-end testing, and automating complex multi-step web workflows without the reliability issues of traditional DOM selectors.

Key Features

01Debugging tools including screenshots, console logs, and network request interception
02Full browser control including navigation, interaction, and state management
03Advanced waiting mechanisms for network idle, text visibility, and JS conditions
040 GitHub stars
05Session management for maintaining authentication and persistent browser states
06Deterministic element selection via accessibility tree snapshots and @refs

Use Cases

01Writing and executing bash-based end-to-end (e2e) tests for web applications
02Debugging and troubleshooting complex UI/UX issues through visual inspection
03Automating multi-step form submissions and data collection workflows