Agent Browser

# Agent Browser Headless browser automation CLI for AI agents featuring deterministic accessibility tree snapshots, session isolation, and state persistence. Agent Browser is a high-performance headless browser automation tool designed specifically for AI agents. It provides a deterministic way to interact with complex web applications by using accessibility tree snapshots and reference-based element selection. This ensures reliable automation even in dynamic single-page applications (SPAs) where traditional selectors might fail. ## How It Works 1 Step 1 Snapshotting Generating an accessibility tree where interactive elements are assigned unique references (e.g., @e1 , @e2 ). 2 Step 2 Interaction Using these stable references to perform actions like click , fill , or hover without needing complex CSS selectors. 3 Step 3 State Management Saving and loading cookies or local storage to maintain authenticated sessions across runs. ## Key Features - **Deterministic Selection**: Uses accessibility tree snapshots with stable refs for reliable element interaction. - **Session Isolation**: Run multiple independent browser sessions simultaneously for complex multi-user workflows. - **State Persistence**: Save and load cookies or local storage to bypass login flows and maintain authentication. - **Network Control**: Intercept, block, or mock network requests to optimize performance or simulate edge cases. - **Comprehensive Interactions**: Supports drag-and-drop, scrolling, iframe switching, and tab management. ## Requirements - **Local Environment**: No external API keys are required. The skill operates using a local Chromium instance managed by the platform. ## Use Cases - **Multi-step Workflows**: Automate complex sequences across different pages or authenticated states. - **Data Extraction**: Navigate to SPAs and extract structured data using JSON-formatted snapshots. - **Parallel Testing**: Use isolated sessions to test user and admin interactions side-by-side in real-time. - **Automated Form Filling**: Efficiently handle complex forms using reference-based element targeting. ## Installation Install via: `npx clawhub@latest install agent-browser-clawdbot`

Instalação

Habilidades relacionadas