vercel-labs/agent-browser

What it solves

agent-browser is a high-performance browser automation CLI designed specifically for AI agents. It removes the complexity of traditional browser automation tools by providing a fast, native Rust-based interface for agents to interact with the web, navigate pages, and extract data without needing a full Node.js or Playwright environment for the daemon.

How it works

The tool operates as a CLI that controls a Chromium-based browser (via Chrome for Testing). It provides a specialized snapshot command that generates an accessibility tree with references (e.g., @e1, @e2), allowing AI agents to identify and interact with elements using stable handles rather than fragile CSS selectors. It also includes a read command to fetch agent-friendly text (Markdown or plain text) from URLs or the active DOM, and a batch mode to execute multiple commands in a single process to reduce overhead.

Who it’s for

Developers building AI agents that need to browse the web, automate browser-based workflows, or perform visual and structural analysis of web pages.

Highlights

• AI-Optimized Interaction: Uses accessibility tree snapshots with element references for more reliable agent navigation.
• Native Performance: Built with Rust for a fast CLI experience.
• Comprehensive Control: Supports full browser interaction including clicking, filling forms, keyboard/mouse control, network interception, and cookie management.
• Agent-Friendly Reading: Specialized read command that prioritizes Markdown and llms.txt for LLM-compatible content extraction.
• Batch Execution: Ability to run multiple commands in one invocation via arguments or JSON stdin to minimize startup latency.
• State Management: Built-in support for saving and loading authentication states and managing browser storage.