molmoweb-mcp-chrome

# molmoweb-mcp-chrome MCP server that exposes [MolmoWeb](https://github.com/allenai/molmoweb) web automation as tools for Claude, using your **real Chrome browser** instead of Playwright. Works via a companion Chrome extension that bridges MCP commands to Chrome APIs. ## Architecture ``` Claude / MCP Client ↓ HTTP (Streamable HTTP transport, port 8002) molmoweb-mcp-chrome (this server) ↕ WebSocket (port 8003) Chrome Bridge Extension (chrome-extension/) ↓ Chrome APIs (tabs, scripting, captureVisibleTab) Active Chrome Tab (your browser) ↓ MolmoWeb API (localhost:8001) ``` ## Key Advantages over Playwright-based MCP | Aspect | molmoweb-mcp (Playwright) | molmoweb-mcp-chrome (this) | |--------|--------------------------|---------------------------| | Browser | Separate Chromium window | Your real Chrome browser | | Sessions | Isolated (no cookies/logins) | Uses your logged-in sessions | | Visibility | Hidden Chromium | See actions in your active tab | | Dependencies | Playwright (~200MB) | ws + express (lightweight) | | Transport | stdio | Streamable HTTP (works with Claude Chrome Extension) | ## Tools | Tool | Description | |------|-------------| | `molmoweb_check_status` | Health check for MolmoWeb backend + Chrome extension connection | | `browser_navigate` | Navigate the active Chrome tab to a URL | | `browser_screenshot` | Capture screenshot of the active Chrome tab | | `browser_get_page_info` | Get current URL and title | | `browser_execute_action` | Execute click/type/scroll/press_key/hover/navigate/wait | | `molmoweb_predict` | Ask MolmoWeb vision model what action to perform | | `run_web_task` | Full autonomous agent loop in your Chrome browser | ## Setup ### 1. Install the MCP server ```bash npm install ``` ### 2. Load the Chrome Bridge Extension 1. Open Chrome → `chrome://extensions/` 2. Enable **Developer Mode** (top right) 3. Click **Load unpacked** → select the `chrome-extension/` directory 4. The extension "MolmoWeb MCP Bridge" should appear and auto-connect when the server runs ### 3. Start the MolmoWeb backend The MolmoWeb vision model must be running at `http://127.0.0.1:8001`: ```bash # On Windows with WSL: run_mcp_server.bat ``` ### 4. Configure in Claude Code Add to your `~/.mcp.json` (global) or project `.mcp.json`: ```json { "mcpServers": { "molmoweb-chrome": { "type": "streamable-http", "url": "http://127.0.0.1:8002/mcp" } } } ``` ### 5. Run the server ```bash npm start ``` Output: ``` [mcp] HTTP server listening on http://127.0.0.1:8002/mcp [ws] WebSocket server listening on ws://127.0.0.1:8003 [mcp] Waiting for Chrome bridge extension to connect... ``` ### Health check ```bash curl http://127.0.0.1:8002/health # {"status":"ok","extension":"connected","sessions":0} ``` ## License MIT