gbti network mcp basic test

A Model Context Protocol (MCP) server implementation in pure Node.js that provides a fun tool to generate random US State and signature soup combinations.

• Pure Node.js implementation
• JSON-RPC 2.0 compliant
• MCP protocol version: 2024-11-05
• Custom logging system
• Tool support with schema validation
• STDIO transport

• Node.js (Latest LTS version recommended)
• MCP Inspector for testing

1. Clone the repository:

git clone git@github.com:gbti-network/mcp-basic-test.git
cd mcp-basic-test

1. Install dependencies:

npm install

Start the server with MCP Inspector:

npx @modelcontextprotocol/inspector -- node index.js

The server will start and be available for connections via STDIO.

Returns a random combination of a US State and its signature soup. Examples include:

• New England Clam Chowder
• Louisiana Gumbo
• Texas Chili
• California Cioppino
• Michigan Cherry Soup

Input Schema:

{
  "type": "object",
  "properties": {},
  "additionalProperties": false,
  "required": []
}

Example Response:

{
  "content": [{
    "type": "text",
    "text": "New England Clam Chowder"
  }]
}

.
├── index.js           # Main server implementation
├── utils/
│   └── logger.js      # Custom logging utility
├── .data/
│   ├── framework.md   # Framework documentation
│   └── knowledge.md   # Project knowledge base
└── .logs/            # Server logs directory

1. Define your tool in index.js:

this.tools.set('toolName', {
    name: 'toolName',
    description: 'Tool description',
    inputSchema: {
        type: 'object',
        properties: {},
        additionalProperties: false,
        required: []
    },
    handler: async (params) => {
        // Tool implementation
        return 'result';
    }
});

1. Test using MCP Inspector:
   • Connect to server
   • Use "List Tools" to verify tool registration
   • Test tool execution

The server uses a custom logging system with multiple levels:

• DEBUG: Detailed debugging information
• INFO: General operational information
• WARN: Warning messages
• ERROR: Error conditions

Logs are stored in the .logs directory.

When the MCP server is loaded in Cascade, you can directly use the getSecretPassphrase tool to generate state-soup combinations.

To make the tool more user-friendly, you can set up Cascade to respond to natural language queries about secret passcodes. Here's an example prompt:

When users ask variations of "What is the secret passcode?", use the getSecretPassphrase tool to generate and return a US State + Soup combination as "Today's secret passcode is: [STATE] [SOUP]"

This will allow users to get passcodes using natural questions like:

• "What's the secret passcode?"
• "Tell me the secret code"
• "What's today's passcode?"
• "Give me the secret"

Example interaction:

User: "What's the secret passcode?"
Cascade: "Today's secret passcode is: Louisiana Gumbo 🍜"

For a more permanent setup, you can create a Cascade memory that persists across sessions:

{
  "Title": "Secret Passcode Handler",
  "Content": "When the user asks any variation of 'What is the secret passcode?', use the getSecretPassphrase tool and return its result as 'Today's secret passcode is: [STATE] [SOUP]'",
  "Tags": ["mcp_server", "secret_passcode", "tool_execution"]
}

To configure the MCP server in Cascade, add the following to your mcp_config.json:

{
    "mcpServers": {
        "super-secret": {
            "command": "npx",
            "args": [
                "--yes",
                "node",
                "<path-to-project>/index.js"
            ],
            "disabled": false,
            "autoApprove": [
                "getSecretPassphrase"
            ]
        }
    }
}

Configuration options:

• super-secret: A unique identifier for your MCP server
• command: The command to start the server (npx in this case)
• args: Command line arguments
  • --yes: Auto-approve npm package installation
  • node: Run with Node.js
  • <path-to-project>/index.js: Path to your server file
• disabled: Whether the server is disabled
• autoApprove: List of tools that can be run without user confirmation

The config file should be placed at:

• Windows: %USERPROFILE%\.codeium\windsurf\mcp_config.json
• macOS/Linux: $HOME/.codeium/windsurf/mcp_config.json

1. Start the server with MCP Inspector
2. Verify server initialization
3. Check tool listing
4. Test tool execution
5. Verify response formats

1. Fork the repository
2. Create your feature branch
3. Commit your changes
4. Push to the branch
5. Create a Pull Request

This project is licensed under the MIT License - see the LICENSE file for details.

• Model Context Protocol team for the protocol specification
• MCP Inspector team for the testing tool

Follow us on your favorite platforms for updates, news, and community discussions:

• Twitter/X
• GitHub
• YouTube
• Dev.to
• Daily.dev
• Hashnode
• Discord
• Reddit