Pair Browsing - AI-Powered Browser Assistant

A Chrome extension that provides an AI-powered assistant to help you navigate and interact with web pages through natural language commands.

Quick Demo

Features

• 🤖 Natural Language Control: Interact with web pages using simple text commands
• 🖱️ Visual Cursor: See where the AI assistant is clicking with an animated cursor
• 🔍 Smart Element Detection: Automatically identifies and interacts with clickable elements
• ⌨️ Form Filling: Fill out forms and input fields with natural language instructions
• 📜 Content Extraction: Extract page content in various formats (text, markdown)
• 🔄 Navigation: Search Google, navigate to URLs, and go back in history
• ⬆️ Scrolling: Control page scrolling with natural commands
• ⚡ Multiple AI Providers: Support for both OpenAI and Google Gemini

Available Actions

1. Click: Click on any interactive element on the page
2. Fill: Input text into form fields
3. Search Google: Perform Google searches directly
4. Navigate: Go to specific URLs or go back in history
5. Scroll: Scroll the page up or down
6. Send Keys: Send keyboard inputs to active elements
7. Extract Content: Get page content as text or markdown

Installation

1. Clone this repository or download the source code
2. Open Chrome and navigate to chrome://extensions/
3. Enable "Developer mode" in the top right corner
4. Click "Load unpacked" and select the directory containing the extension files

Configuration

1. Click the extension icon in Chrome to open the options page
2. Choose your preferred AI provider (OpenAI or Google Gemini)
3. Configure the API settings:

For OpenAI:

• Enter your OpenAI API key
• Optionally customize the model (default: gpt-4o-mini)

For Google Gemini:

• Enter your Gemini API key
• Optionally customize the model (default: gemini-2.0-flash-exp)

Usage

1. Click the extension icon to open the sidebar
2. Type your command in natural language (e.g., "Click the login button" or "Fill the email field with [email protected]")
3. The AI assistant will:
   • Analyze the page structure
   • Identify relevant elements
   • Execute the requested action
   • Provide visual feedback with the cursor

Debug Mode

Enable debug mode in the options page to:

• View captured screenshots in the chat
• See detailed logging information
• Help troubleshoot interactions

Examples

Here are some example commands you can try:

• "Click the sign up button"
• "Fill the username field with johndoe"
• "Search for Chrome extensions"
• "Scroll down one page"
• "Go back to the previous page"
• "Extract the main content as markdown"

Requirements

• Google Chrome browser
• API key from either OpenAI or Google Gemini
• Active internet connection

Technical Details

The extension consists of:

• Content Script: Handles page interactions and element detection
• Background Script: Manages AI communication and extension state
• Sidebar Interface: Provides the chat interface for user commands
• Options Page: Allows configuration of AI providers and settings

Privacy & Security

• API keys are stored locally in Chrome storage
• Screenshots are only used for AI analysis and are not stored
• No data is collected or stored outside of your browser

Troubleshooting

If the extension isn't working as expected:

1. Check if your API key is correctly configured
2. Ensure you have an active internet connection
3. Try refreshing the page
4. Check the browser console for error messages
5. Enable debug mode for more detailed logging

License

This project is licensed under the MIT License - see the LICENSE file for details.