eru-ai MCP Bridge

A bridge that connects MCP clients (Claude Desktop, Cursor) to the eru-ai WebSocket server via stdio interface.

Installation

Global Installation

npm install -g @eru-tech/eru-ai-mcp-bridge

Using npx (Recommended)

npx @eru-tech/eru-ai-mcp-bridge

Usage

With Default Settings

npx @eru-tech/eru-ai-mcp-bridge

This connects to ws://localhost:8088/mcp/websocket by default.

With Custom Server URL

npx @eru-tech/eru-ai-mcp-bridge --server ws://your-server:8088/mcp/websocket

With Verbose Logging

npx @eru-tech/eru-ai-mcp-bridge --verbose

Configuration for MCP Clients

Claude Desktop

Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "eru-ai": {
      "command": "npx",
      "args": ["@eru-tech/eru-ai-mcp-bridge", "--server", "ws://your-server:8088/mcp/websocket"],
      "env": {}
    }
  }
}

Cursor IDE

Add to your .cursor/mcp.json or global MCP configuration:

{
  "mcpServers": {
    "eru-ai": {
      "command": "npx",
      "args": ["@eru-tech/eru-ai-mcp-bridge", "--server", "ws://your-server:8088/mcp/websocket"]
    }
  }
}

Features

• ✅ WebSocket to stdio bridge: Converts WebSocket MCP protocol to stdio interface
• ✅ Remote server support: Connect to any eru-ai server over the network
• ✅ Zero configuration: Works out of the box with default eru-ai server
• ✅ Custom server URLs: Support for different server endpoints
• ✅ Error handling: Graceful error handling and connection management
• ✅ npx support: Can be run directly with npx without installation

How It Works

1. MCP Client (Claude Desktop/Cursor) launches the bridge via stdio
2. Bridge connects to eru-ai WebSocket server
3. Messages are translated between stdio (JSON-RPC) and WebSocket protocols
4. Tools from eru-ai server are available in your MCP client

eru-ai Server Requirements

Your eru-ai server must:

• Be running on the specified port (default: 8088)
• Have MCP WebSocket endpoint enabled at /mcp/websocket
• Be accessible from where the bridge is running

Troubleshooting

Connection Refused

npx @eru-tech/eru-ai-mcp-bridge --verbose

Check that:

• eru-ai server is running
• Port 8088 is accessible
• WebSocket endpoint /mcp/websocket is available

MCP Client Not Finding Tools

Verify the server URL and check eru-ai server logs:

curl http://localhost:8088/mcp/health
curl http://localhost:8088/mcp/stats

Development

git clone <repo>
cd eru-ai-mcp-bridge
npm install
node index.js --verbose

License

MIT - See LICENSE file for details.