Index Network MCP Server

A Model Context Protocol (MCP) server that enables ChatGPT integration with Index Network's discovery protocol.

Features

• 🚀 Quick Setup: One-command startup with included script
• 🔧 MCP Tools: Health check, echo, and search functionality
• 🌐 Public Access: Built-in ngrok integration for external access
• 🧪 Testing: MCP Inspector integration for easy testing

Quick Start

Prerequisites

• Node.js 18+
• ngrok account (sign up free)

Installation

# Clone and setup
git clone https://github.com/indexnetwork/mcp.git
cd mcp
npm install

# Configure environment
cp env.example .env
# Edit .env with your ngrok authtoken

Run Everything at Once

# Start server, ngrok tunnel, and inspector
./src/start-mcp.sh

The script will:

• Start the MCP server on port 3002
• Create an ngrok tunnel (public URL)
• Launch MCP Inspector for testing
• Display all URLs and process IDs

Manual Setup (Alternative)

# Terminal 1: Start server
npm run dev

# Terminal 2: Create tunnel
ngrok http 3002

# Terminal 3: Test with inspector
npx @modelcontextprotocol/inspector@latest
# Enter: http://localhost:3002/mcp

Configuration

Environment Variables

Variable	Description	Default
MCP_SERVER_PORT	Server port	3002
NGROK_AUTHTOKEN	ngrok auth token	Required
NODE_ENV	Environment	development

ngrok Setup

1. Get your authtoken from ngrok dashboard
2. Add to .env:

   NGROK_AUTHTOKEN=your_token_here
   

Available Tools

health-check

Check server status.

echo

Echo back a message.

search

Search functionality (demo implementation).

ChatGPT Integration

1. Run the startup script: ./src/start-mcp.sh
2. Copy the ngrok URL from the output (e.g., https://abc123.ngrok-free.dev/mcp)
3. Configure ChatGPT to use this MCP server URL

Development

Project Structure

mcp/
├── src/
│   ├── server.ts          # MCP server implementation
│   └── start-mcp.sh       # Startup script
├── docs/                  # Documentation
├── package.json           # Dependencies
└── README.md             # This file

Adding Tools

server.registerTool("my-tool", {
  title: "My Tool",
  description: "Tool description",
  inputSchema: z.object({
    input: z.string()
  })
}, async (params) => ({
  content: [{ type: "text", text: "Result" }],
  _meta: { "openai/widgetAccessible": true }
}));

Building

npm run build    # Compile TypeScript
npm start        # Run production build

Troubleshooting

Script won't run

• Make executable: chmod +x src/start-mcp.sh
• Check Node.js version: node --version (needs 18+)

ngrok errors

• Verify authtoken in .env
• Check if port 3002 is available

MCP Inspector issues

• Ensure server is running first
• Use correct URL: http://localhost:3002/mcp

Links

• Model Context Protocol
• Index Network
• ngrok

License

MIT License - see LICENSE for details.