vercel-labs-agent-browser

Overview

Agent Browser is a headless browser automation CLI designed specifically for AI agents. Built with a fast Rust core and a Node.js fallback, it provides a comprehensive set of commands for navigating, clicking, filling forms, taking screenshots, and extracting data from web pages. Its accessibility-tree-based snapshot system with element refs makes it particularly well-suited for AI-driven browser interactions.

Key Features

• Accessibility tree snapshots — The snapshot command returns a structured accessibility tree with element refs, enabling AI agents to interact with page elements by reference rather than fragile CSS selectors.
• Comprehensive browser commands — Supports clicking, typing, filling, scrolling, drag-and-drop, file uploads, screenshots, PDF export, and JavaScript evaluation all from the command line.
• Dual selector support — Works with both ref-based selectors from snapshots (e.g., @e2) and traditional CSS selectors (e.g., #submit), giving flexibility in how elements are targeted.
• Session and tab management — Supports multiple pages, persistent sessions via --profile, and CDP connections, allowing complex multi-page workflows.
• Fast Rust CLI — Ships with a native Rust binary for performance, with automatic fallback to Node.js if needed.

Who is this for?

Agent Browser is built for developers and AI engineers who need reliable, programmatic browser automation for AI agent workflows. It is especially useful for teams building autonomous agents that need to interact with web applications, run end-to-end tests, or scrape structured data from websites.