Browser Automation

Control web browsers with natural language commands using Playwright

What You'll Learn

Navigate websites and interact with elements

Take screenshots (full page or viewport)

Extract content and scrape data

Control browser with natural language

Quick Start

browser_example.py

from connectonion import Agent
from connectonion.tools import BrowserTool

# Initialize browser tool
browser = BrowserTool()

# Create agent with browser capabilities
agent = Agent("browser-bot", tools=[browser])

# Control browser with natural language
result = agent.input("Navigate to example.com and take a screenshot")
print(result)

# Extract content
content = agent.input("Extract all the links from the page")
print(content)

# Complex interactions
agent.input("Fill the search box with 'AI agents' and click search")

Python REPL

Interactive

Screenshot saved to example_screenshot.png

[{text: 'About', href: '/about'}, {text: 'Contact', href: '/contact'}]

Search completed successfully

Complete Browser Agent

agent.py

#!/usr/bin/env python3
"""Browser automation agent with Playwright integration"""

import os
from connectonion import Agent
from playwright.sync_api import sync_playwright
from typing import Optional, Dict, Any

class BrowserTool:
    """Tool for browser automation using Playwright"""
    
    def __init__(self):
        self.playwright = None
        self.browser = None
        self.page = None
        
    def start_browser(self, headless: bool = False) -> str:
        """Start a new browser instance"""
        self.playwright = sync_playwright().start()
        self.browser = self.playwright.chromium.launch(headless=headless)
        self.page = self.browser.new_page()
        return "Browser started successfully"
    
    def navigate(self, url: str) -> str:
        """Navigate to a URL"""
        if not self.page:
            return "Browser not started. Please start browser first."
        self.page.goto(url)
        return f"Navigated to {url}"
    
    def screenshot(self, filename: str = "screenshot.png", full_page: bool = False) -> str:
        """Take a screenshot of the current page"""
        if not self.page:
            return "Browser not started. Please start browser first."
        self.page.screenshot(path=filename, full_page=full_page)
        return f"Screenshot saved to {filename}"
    
    def extract_text(self, selector: str = "body") -> str:
        """Extract text content from the page"""
        if not self.page:
            return "Browser not started. Please start browser first."
        element = self.page.query_selector(selector)
        if element:
            return element.text_content()
        return "No content found"
    
    def click(self, selector: str) -> str:
        """Click an element on the page"""
        if not self.page:
            return "Browser not started. Please start browser first."
        self.page.click(selector)
        return f"Clicked element: {selector}"
    
    def fill(self, selector: str, text: str) -> str:
        """Fill a form field with text"""
        if not self.page:
            return "Browser not started. Please start browser first."
        self.page.fill(selector, text)
        return f"Filled {selector} with text"
    
    def extract_links(self) -> list:
        """Extract all links from the current page"""
        if not self.page:
            return []
        links = self.page.eval_on_selector_all(
            "a[href]", 
            "elements => elements.map(e => ({text: e.textContent, href: e.href}))"
        )
        return links
    
    def close_browser(self) -> str:
        """Close the browser and clean up"""
        if self.browser:
            self.browser.close()
        if self.playwright:
            self.playwright.stop()
        return "Browser closed"

# Create the browser tool instance
browser = BrowserTool()

# Create agent with browser tool
agent = Agent(
    "browser-agent",
    tools=[browser],
    system_prompt="""You are a browser automation assistant.
    Help users navigate websites, take screenshots, and extract content.
    Always start the browser before performing actions.
    Be helpful and explain what you're doing."""
)

if __name__ == "__main__":
    # Example usage
    print("🌐 Browser Automation Agent")
    print("=" * 50)
    
    # Start browser
    result = agent.input("Start the browser in headless mode")
    print(f"✅ {result}")
    
    # Navigate to a website
    result = agent.input("Navigate to https://example.com")
    print(f"✅ {result}")
    
    # Take screenshot
    result = agent.input("Take a full page screenshot and save as example.png")
    print(f"✅ {result}")
    
    # Extract content
    result = agent.input("Extract all the links from the page")
    print(f"📋 Links found: {result}")
    
    # Clean up
    result = agent.input("Close the browser")
    print(f"✅ {result}")

Interactive Demo Script

demo.py

#!/usr/bin/env python3
"""Interactive demo of the browser agent"""

from agent import agent, browser

print("🌐 Browser Automation Demo")
print("=" * 50)

# Step 1: Start browser
print("Step 1: Starting browser...")
result = agent.input("Start the browser (not headless so we can see it)")
print(f"✅ {result}\n")

# Step 2: Navigate to documentation
print("Step 2: Navigating to ConnectOnion docs...")
result = agent.input(
    "Navigate to https://docs.connectonion.com and tell me the page title"
)
print(f"✅ {result}\n")

# Step 3: Take screenshot
print("Step 3: Taking a screenshot...")
result = agent.input("Take a screenshot and save it as docs_homepage.png")
print(f"✅ {result}\n")

# Step 4: Extract navigation links
print("Step 4: Extracting navigation links...")
result = agent.input(
    "Extract all navigation links and tell me what sections are available"
)
print(f"📋 {result}\n")

# Step 5: Navigate to examples
print("Step 5: Going to examples section...")
result = agent.input(
    "Click on the Examples link if available and tell me what you see"
)
print(f"✅ {result}\n")

# Step 6: Full page screenshot
print("Step 6: Taking full page screenshot...")
result = agent.input(
    "Take a full page screenshot of the examples and save as examples_full.png"
)
print(f"✅ {result}\n")

# Clean up
print("Cleaning up...")
result = agent.input("Close the browser")
print(f"✅ {result}")

print("\n" + "=" * 50)
print("Demo complete! Check out:")
print("  - docs_homepage.png")
print("  - examples_full.png")

Python REPL

Interactive

🌐 Browser Automation Demo

==================================================

Step 1: Starting browser...

✅ Browser started successfully

Step 2: Navigating to ConnectOnion docs...

✅ Navigated to https://docs.connectonion.com - Title: "ConnectOnion Documentation"

Step 3: Taking a screenshot...

✅ Screenshot saved to docs_homepage.png

Step 4: Extracting navigation links...

📋 Found 12 sections: Getting Started, Core Concepts, Advanced Features, Examples, Blog, Roadmap

Step 5: Going to examples section...

✅ Clicked Examples link - Now viewing example projects

Step 6: Taking full page screenshot...

✅ Full page screenshot saved to examples_full.png

Cleaning up...

✅ Browser closed

==================================================

Demo complete! Check out:

- docs_homepage.png

- examples_full.png

Common Use Cases

Website Monitoring

main.py

# Monitor website changes
agent.input("Navigate to status.example.com")
agent.input("Take a screenshot and save with timestamp")
agent.input("Extract the status text and check if all systems operational")

Python REPL

Interactive

Navigated to status.example.com

Screenshot saved: status_20250906_150000.png

Status: All systems operational ✅

Data Extraction

main.py

# Scrape product information
agent.input("Navigate to shop.example.com/products")
agent.input("Extract all product names and prices")
agent.input("Save the data to products.json")

Python REPL

Interactive

Navigated to shop.example.com/products

Extracted 25 products with prices

Data saved to products.json

Form Automation

main.py

# Fill and submit forms
agent.input("Navigate to example.com/contact")
agent.input("Fill the name field with 'John Doe'")
agent.input("Fill the email field with 'john@example.com'")
agent.input("Click the submit button")

Python REPL

Interactive

Navigated to example.com/contact

Filled name field

Filled email field

Form submitted successfully

Installation

Install dependencies

$pip install connectonion

$pip install playwright

$playwright install

Pro Tips

•Use headless mode for production to save resources
•Add waits for dynamic content: page.wait_for_selector()
•Handle errors gracefully with try-except blocks
•Use specific selectors for reliable element targeting
•Clean up resources with browser.close()

Ready to Automate the Web?

Start building your own browser automation agents with ConnectOnion

View All Examples View on GitHub

Calculator

All Posts