agent-browser.dev
agent-browser is a command-line interface for headless browser automation, designed for AI agents. It uses a Rust CLI and a Node.js daemon that manages a Playwright browser, and outputs compact text snapshots with element references.
Visit website (opens in a new tab)The Headless Browser Automation CLI, referred to as agent-browser, is a specialized tool designed for AI agents. This command-line interface (CLI) focuses on efficient interactions with web browsers using a unique architecture that minimizes the context usage required by AI systems. By employing a fast Rust-based implementation with support for Node.js as a fallback, agent-browser significantly enhances the browsing experience for AI entities.
To begin using agent-browser, users can install it via command-line interfaces:
For all platforms, including Windows, Linux, and macOS, the fastest native Rust CLI can be installed using npm:
npm install -g agent-browser
Alternatively, for macOS users, it can be installed using Homebrew:
brew install agent-browser
Those who prefer not to install can try the utility directly with npx:
npx agent-browser open example.com
agent-browser comes equipped with a variety of features tailored for optimizing browser automation for AI:
The tool prioritizes an agent-first approach, providing compact text output that consumes fewer tokens than traditional JSON formats, thereby enhancing context efficiency for AI.
The snapshot command generates an accessibility tree with unique references (refs) for elements. This allows for deterministic and efficient interactions with specific elements on the page.
With a Rust-based CLI, command parsing is executed quickly, ensuring that users experience minimal latency when issuing commands.
The CLI supports over 50 commands covering a range of functionalities, such as navigation, form submissions, screenshot capturing, network interactions, and storage manipulation.
Users can run multiple isolated browser instances, each with separate authentication credentials, facilitating parallel operations without overlap.
The tool supports multiple operating systems including macOS, Linux, and Windows, with native binaries ensuring optimal performance on each platform.
The agent-browser is designed to integrate seamlessly with various AI systems, including:
It works with any agent capable of executing shell commands, making it a versatile choice for developers utilizing AI in their projects.
To illustrate its functionality, here are some example commands:
Navigate to a website and capture a snapshot:
agent-browser open example.com
agent-browser snapshot -i
Sample Output:
- heading "Example Domain" [ref=e1]
- link "More information..." [ref=e2]
Interact with the identified elements using their references:
agent-browser click @e2
agent-browser screenshot page.png
agent-browser close
The unique ref-based system enhances the experience in several key ways:
The agent-browser employs an innovative client-daemon architecture to maximize performance:
This daemon starts automatically and persists, making it available between commands without needing to be re-invoked after every interaction.
agent-browser provides native binaries for the following platforms:
The agent-browser is a powerful CLI tool for browser automation within AI contexts. Its compact and efficient design allows developers and AI practitioners to interact with web applications more effectively, streamlining tasks and reducing overhead associated with traditional methods. This tool is particularly beneficial for developers integrating AI systems with web technologies, ensuring they can harness the full power of AI while maintaining optimal performance and efficiency. Whether for personal projects or professional applications, agent-browser offers a unique and valuable solution for headless browser automation.
Ready to try it? Visit agent-browser.dev (opens in a new tab)
