buse - Natural Language Browser Automation

The buse module enables users to control a web browser using natural language through the browser-use framework. It automates complex browser actions without requiring manual coding.

When to Activate

• When the user wants an AI to navigate websites or interact with web elements.
• When performing automated browser tasks like documentation search or visual verification (screenshots).
• When using a terminal UI to interactively guide a browser agent.
• When running headless browser tasks for data extraction or testing.

Core Principles & Rules

• Safety & Permissions: AI-generated browser actions should be monitored; use caution with sensitive sites.
• Model Requirement: Requires an AI model API key (OpenAI, Gemini, etc.) configured via respective x-cmd modules.
• Context Handling: Supports custom browser windows, user data directories, and CDP connections.

Additional Scenarios

• Visual Evidence: Automatically take screenshots during a browser task using specific prompts.
• MCP Mode: Run as a Model Context Protocol server for integration with other AI tools.

Patterns & Examples

Interactive Browser Control

# Launch the interactive terminal UI for browser control
x buse

Direct Task Execution

# Run a specific browser task non-interactively
x buse -p "Search for OpenAI documentation and take a screenshot of the homepage"

Headless Mode

# Run a task in the background without a visible window
x buse --headless -p "Check the stock status of an item on ExampleStore.com"

Checklist

• Ensure browser-use and Chromium are installed via x buse --install.
• Confirm that at least one AI model API key is set up.
• Verify if the task requires a visible window or can run headless.