Browser Use

Browser Use 2.0 — Fast CDP-Based Browser Automation

Browser Use 2.0 CLI — direct CDP connection to Chrome, 2x speed, half the cost.

Quick Start

# Install
pip3 install browser-use

# Basic session
browser-use open https://example.com
browser-use state                    # See clickable elements with indices
browser-use click 5                  # Click element by index
browser-use type "Hello"             # Type text
browser-use screenshot page.png
browser-use close

When to Use

Use browser-use when:

• You need to connect to an already-running Chrome instance via CDP
• Speed and cost efficiency are priorities
• You want direct CDP protocol access

Use playwright-cli instead for:

• General browser automation (default choice)
• Complex element targeting with refs
• Tracing, video recording, network mocking

Connect to Running Chrome

# Launch Chrome with remote debugging
/Applications/Google\ Chrome.app/Contents/MacOS/Google\ Chrome --remote-debugging-port=9222

# Connect browser-use to it
browser-use connect --cdp-url http://localhost:9222

This is the key advantage — connect to a Chrome session where the user is already logged in, preserving all auth state.

Commands

Core

browser-use open https://example.com
browser-use state                    # List interactive elements
browser-use click <index>            # Click by element index
browser-use type "text"              # Type text
browser-use press Enter              # Keyboard input
browser-use screenshot [filename]
browser-use close

CDP Connection

browser-use connect --cdp-url http://localhost:9222
browser-use connect --cdp-url ws://localhost:9222/devtools/browser/<id>

State Inspection

browser-use state                    # All interactive elements
browser-use state --full             # Full page state

Python API (for scripts)

from browser_use import Agent, Browser, BrowserConfig

# Connect to existing Chrome
config = BrowserConfig(cdp_url="http://localhost:9222")
browser = Browser(config=config)

# Or launch new browser
browser = Browser()

agent = Agent(
    task="Find the latest news about AI",
    llm=your_llm,
    browser=browser,
)
result = await agent.run()

Environment Variables

• BROWSER_USE_API_KEY — Optional, for cloud stealth browsers
• LLM provider keys (if using Agent mode): OPENAI_API_KEY, ANTHROPIC_API_KEY, etc.

Integration with api-harvester

browser-use's CDP connection is useful for api-harvester Mode B:

1. Connect to user's logged-in Chrome session
2. Capture network traffic while user operates the site
3. Extract auth tokens from the live session