opendia

OpenDia provides a browser automation and AI-driven control layer that lets models interact with your browser using your existing sessions, data, and extensions. This MCP server exposes a set of browser tools that enable AI to analyze pages, click elements, fill forms, navigate across tabs, and manage complex web apps—without requiring you to start from scratch. You can run OpenDia locally with a single command and connect it to your preferred MCP-enabled assistant (such as Claude Desktop or Cursor) to start automating tasks like social media posting, research, testing workflows, and more. The system emphasizes privacy and local-first operation, leveraging your existing browser state, cookies, saved passwords, and extensions to carry out real-world actions securely on your device.

To use OpenDia with MCP tools, install and start the server (for example via npx opendia), then configure your MCP client with the opendia entry shown in the Quick Start snippet. The client communicates with the server over WebSocket and HTTP endpoints (default ports can be customized). Use the client’s command or the provided interface to initiate tasks, leveraging OpenDia’s capabilities such as smart page understanding, natural interactions (clicking, filling, navigating), and tab/window management to automate workflows across sites and apps.

Prerequisites:\n- Node.js and npm installed on your machine (Node.js 14+ recommended).\n- Access to a shell/terminal.\n- Optional: ngrok if you plan to use Auto-Tunnel mode.\n\nInstallation steps:\n1) Ensure Node.js and npm are installed. You can verify with:\n node --version\n npm --version\n\n2) Install or run the OpenDia package via npx (no global install required):\n npx opendia\n\n If you prefer to point to a specific version and avoid prompts, you can also run:\n npx -y opendia\n\n3) Connect OpenDia to your MCP client:\n Use the following configuration in your MCP client (Claude Desktop or Cursor):\n {\n "mcpServers": {\n "opendia": {\n "command": "npx",\n "args": ["opendia"]\n }\n }\n }\n\n4) (Optional) If you plan to use Auto-Tunnel (ngrok) for remote access, install and configure ngrok as described in the README, including adding your authtoken.\n\n5) Start the server and test locally:\n - Local mode (default): npx opendia\n - Verify connections: ws://localhost:5555 and http://localhost:5556 (or your configured ports).\n\nPrereq notes:\n- Ensure your browser extension(s) required by OpenDia are installed and enabled (Chrome/Chromium or Firefox).\n- Some features rely on the browser extension being active and on the availability of an active user session for the target sites.

Tips and common issues:\n- Privacy and security: OpenDia leverages your local browser data (sessions, cookies, passwords) and runs locally. Ensure your system security settings allow the extension to operate with your accounts.\n- Port configuration: You can customize WebSocket/HTTP ports as needed. If you have port conflicts, the OpenDia startup will terminate existing OpenDia processes to reclaim ports.\n- Auto-tunnel caveat: Ngrok is required for auto-tunneling. Install ngrok per your OS and configure tokens as described in the README.\n- Browser compatibility: OpenDia supports Chrome, Firefox, and any Chromium-based browser. If any site blocks automation, use the anti-detection and compatibility features described in the Capabilities section.\n- Troubleshooting: If you encounter connection issues, verify that the MCP client is configured with the opendia server entry, and check that the server process started successfully via npx opendia.\n- Updates: When updating, you can rely on npx to fetch the latest OpenDia package from npm without a global install.