mirror of
https://github.com/aaronjmars/opendia.git
synced 2025-12-29 16:16:00 +00:00
204 lines
7.5 KiB
Markdown
204 lines
7.5 KiB
Markdown
# OpenDia ✳️
|
|
|
|
OpenDia is an open alternative to Dia. Connect to your browser with MCP & do anything.
|
|
> Exposes browser functions through the Model Context Protocol (MCP), allowing AI models to interact with browser capabilities.
|
|
|
|
## ⚠️ Security Warning
|
|
|
|
**IMPORTANT**: This extension is provided as-is with no security guarantees. By using this extension, you acknowledge and accept the following risks:
|
|
|
|
- The extension requires broad browser permissions to function
|
|
- It establishes WebSocket connections to localhost
|
|
- It allows external applications to control browser functions
|
|
- We cannot guarantee the security of data transmitted through the extension
|
|
- Use at your own risk and only in trusted environments
|
|
|
|
## Quick Start
|
|
|
|
### Prerequisites
|
|
- Node.js (v14 or higher)
|
|
- Google Chrome browser
|
|
|
|
### Installation
|
|
|
|
1. **Set up the MCP Server**
|
|
```bash
|
|
cd opendia-mcp
|
|
npm install
|
|
npm start
|
|
```
|
|
|
|
2. **Install the Chrome Extension**
|
|
- Open Chrome and go to `chrome://extensions/`
|
|
- Enable "Developer mode" in the top right
|
|
- Click "Load unpacked" and select the `opendia-extension` directory
|
|
- The extension icon will appear in your browser toolbar
|
|
|
|
3. **Configure your MCP client**
|
|
Add the browser server to your MCP configuration:
|
|
```json
|
|
{
|
|
"mcpServers": {
|
|
"opendia": {
|
|
"command": "node",
|
|
"args": ["/path/to/opendia/opendia-mcp/server.js"],
|
|
"env": {}
|
|
}
|
|
}
|
|
}
|
|
```
|
|
|
|
## Enhanced MCP Tools (17 Total)
|
|
|
|
### 🌐 Web Browser Automation Tools (8 Tools)
|
|
|
|
- **page_analyze**: Two-phase intelligent page analysis with anti-detection bypass
|
|
- Phase 1 (discover): Quick scan with element state detection
|
|
- Phase 2 (detailed): Full analysis with element fingerprinting
|
|
- Enhanced pattern database with confidence scoring
|
|
- Supports Twitter/X, GitHub, and universal patterns
|
|
|
|
- **page_extract_content**: Smart content extraction with summarization
|
|
- Extract articles, search results, or social media posts
|
|
- Token-efficient summaries with quality metrics
|
|
- Site-specific extraction patterns for Twitter/X, GitHub, Google
|
|
|
|
- **element_click**: Reliable element clicking with smart targeting
|
|
- Uses element IDs from page analysis
|
|
- Supports different click types (left, right, double)
|
|
- Auto-scrolls elements into view
|
|
|
|
- **element_fill**: Enhanced form filling with anti-detection bypass
|
|
- Specialized bypasses for Twitter/X, LinkedIn, Facebook
|
|
- Natural focus sequence: click → focus → fill
|
|
- Comprehensive event simulation for modern web apps
|
|
|
|
- **element_get_state**: Get detailed element state information
|
|
- Check if elements are disabled, clickable, visible
|
|
- Get current values and element properties
|
|
- Essential for conditional automation logic
|
|
|
|
- **page_navigate**: Enhanced navigation with wait conditions
|
|
- Navigate to URLs with optional element wait conditions
|
|
- Timeout handling and error reporting
|
|
|
|
- **page_wait_for**: Conditional waiting for elements or text
|
|
- Wait for elements to become visible
|
|
- Wait for specific text to appear on page
|
|
- Configurable timeout periods
|
|
|
|
- **page_scroll**: Scroll pages in various directions
|
|
- Critical for long pages and infinite scroll content
|
|
- Supports smooth scrolling and element targeting
|
|
|
|
### 📑 Tab Management Tools (4 Tools)
|
|
|
|
- **tab_create**: Create new tabs with advanced options
|
|
- Create tabs with or without URLs
|
|
- Control tab activation and focus
|
|
- Wait for elements to load after creation
|
|
- Perfect for multi-tab workflows
|
|
|
|
- **tab_close**: Close tabs with flexible targeting
|
|
- Close current tab, specific tab by ID, or multiple tabs
|
|
- Batch close operations for cleanup
|
|
- Safe handling of tab closure
|
|
|
|
- **tab_list**: Get comprehensive tab information
|
|
- List all open tabs with details (title, URL, status)
|
|
- Filter by current window or all windows
|
|
- Track active tab and tab states
|
|
|
|
- **tab_switch**: Switch between tabs intelligently
|
|
- Switch to specific tabs by ID
|
|
- Focus windows automatically
|
|
- Essential for multi-tab automation workflows
|
|
|
|
### 📊 Browser Data Access Tools (5 Tools)
|
|
|
|
- **get_bookmarks**: Get all bookmarks or search for specific ones
|
|
- Search bookmarks by query string
|
|
- Returns structured bookmark data with hierarchy
|
|
|
|
- **add_bookmark**: Add new bookmarks
|
|
- Create bookmarks with title and URL
|
|
- Optional parent folder support for organization
|
|
|
|
- **get_history**: Search browser history with comprehensive filters
|
|
- Advanced filtering by date, domains, visit count, keywords
|
|
- Sophisticated sorting and metadata extraction
|
|
- Perfect for finding previous work and research
|
|
|
|
- **get_selected_text**: Get currently selected text on the page
|
|
- Rich metadata about selection context and position
|
|
- Includes parent element information and page context
|
|
- Configurable length limits and truncation
|
|
|
|
- **get_page_links**: Get all hyperlinks on current page with filtering
|
|
- Smart filtering for internal/external links
|
|
- Domain-specific filtering options
|
|
- Essential for link analysis and navigation planning
|
|
|
|
## 🚀 Key Features
|
|
|
|
### Hybrid Intelligence Architecture
|
|
- **99% Local Operations**: Pattern database eliminates most LLM calls ($0 cost vs $20+/month)
|
|
- **Pattern Database**: Pre-built selectors for Twitter/X, GitHub, and common patterns
|
|
- **Semantic Analysis**: Fallback using HTML semantics and ARIA labels
|
|
- **Confidence Scoring**: Reliable element detection with quality metrics
|
|
|
|
### Visual Testing Interface
|
|
- **Real-time Testing**: Test content extraction and page analysis
|
|
- **Element Highlighting**: Visual feedback with confidence-based colors
|
|
- **Performance Metrics**: Execution time and data size monitoring
|
|
- **JSON Viewer**: Full result inspection and debugging
|
|
|
|
## Project Structure
|
|
|
|
```
|
|
opendia/
|
|
├── opendia-extension/ # Chrome extension files
|
|
│ ├── manifest.json # Extension configuration
|
|
│ ├── background.js # Background service worker
|
|
│ ├── content.js # Content scripts
|
|
│ ├── popup.html # Extension popup UI
|
|
│ └── popup.js # Popup functionality
|
|
├── opendia-mcp/ # MCP server implementation
|
|
│ ├── package.json # Server dependencies
|
|
│ ├── server.js # MCP server logic
|
|
│ └── .env # Environment configuration
|
|
└── README.md
|
|
```
|
|
|
|
## Contributing
|
|
|
|
1. **Adding New Browser Functions**
|
|
- Add the tool definition in `getAvailableTools()` in background.js
|
|
- Implement the handler in `handleMCPRequest()`
|
|
- The tool will automatically be registered with the MCP server
|
|
|
|
2. **Development Workflow**
|
|
- Modify extension files in the `opendia-extension` directory
|
|
- Reload the extension in Chrome to see changes
|
|
- Test new functionality through the MCP interface
|
|
|
|
3. **Security Considerations**
|
|
- Review and limit permissions based on needs
|
|
- The WebSocket server runs on localhost only by default
|
|
- Be cautious when executing scripts or accessing sensitive data
|
|
- Consider adding authentication between the extension and MCP server
|
|
|
|
## Troubleshooting
|
|
|
|
- **Extension not connecting**: Check that the MCP server is running on port 3000
|
|
- **Tools not available**: Verify the extension is loaded and check the popup for connection status
|
|
- **Permission errors**: Ensure the extension has the necessary permissions in manifest.json
|
|
|
|
## License
|
|
|
|
MIT License
|
|
|
|
## Disclaimer
|
|
|
|
This software is provided "as is", without warranty of any kind, express or implied. In no event shall the authors or copyright holders be liable for any claim, damages or other liability arising from the use of this software.
|