MCP Server

MCP Server

A robust server implementation of the Model Context Protocol (MCP) that supports multiple transport methods (SSE and STDIO) for real-time communication with AI models.

🚀 Features

• Full MCP Implementation: Complete and standard-compliant implementation of the Model Context Protocol
• Multiple Transports: Seamless communication via SSE (Server-Sent Events) and STDIO
• Robust Architecture: Built on Clean Architecture and SOLID principles for maintainability and extensibility
• Powerful Tooling: Easily extensible tools, resources, and prompts
• Developer-Friendly: Comprehensive logging, error handling, and testing support

🏗️ Architecture

The project follows a clean, layered architecture adhering to SOLID principles:

Layer	Responsibility	Components
Domain	Core business logic	Interfaces, Models, Entities
Application	Use cases, application flow	Services, Use Cases, DTOs
Infrastructure	Technical concerns	Implementations, Adapters, External services
Presentation	User interface	API Controllers, HTTP endpoints

Architectural Diagram

┌─────────────────────┐      ┌─────────────────────┐
│  Presentation       │      │  Application        │
│  Layer              │      │  Layer              │
│  - Express API      │─────▶│  - Use Cases        │
│  - Controllers      │      │  - Services         │
└─────────────────────┘      └──────────┬──────────┘
                                        │
                                        ▼
┌─────────────────────┐      ┌─────────────────────┐
│  Infrastructure     │      │  Domain             │
│  Layer              │◀─────│  Layer              │
│  - Implementations  │      │  - Interfaces       │
│  - Adapters         │      │  - Models           │
└─────────────────────┘      └─────────────────────┘

💻 Requirements

• Node.js 18 or higher
• Yarn or npm package manager

📦 Installation

# Clone the repository
git clone https://github.com/your-username/mcp-server.git
cd mcp-server

# Install dependencies
yarn install

# Build the project
yarn build

🚀 Running the Server

# Start the server
yarn start

# Build and start in one command (for development)
yarn rebuild

By default, the server runs on port 3001. You can configure this using the PORT environment variable.

🔌 Transport Options

The MCP server supports two transport methods:

Transport	Description	Use Case
SSE	Server-Sent Events over HTTP	Web applications, browser clients
STDIO	Standard input/output streams	CLI tools, local applications

🛣️ API Endpoints

• SSE: /sse - Establish SSE connections for real-time communication
• Messages: /messages - Send JSON-RPC messages to the server

🧪 Testing

Use the official MCP Inspector tool to test your server implementation:

1. Build and start your MCP Server

yarn build
yarn rebuild  # if using SSE transport (builds & starts)
# OR
yarn start    # manual start

2. Run the inspector against your server

npx @modelcontextprotocol/inspector node build/index.js

3. Access to Inspecto UI http://127.0.0.1:6274 🚀

🧩 Extending Functionality

Adding a New Tool

// tools/custom/my-tool.ts
import { z } from "zod";
import { McpServer } from "@modelcontextprotocol/sdk/server/mcp.js";
import { ITool } from "../core/domain/interfaces/tool.interface.js";
import { ILogger } from "../core/domain/interfaces/logger.interface.js";

export class MyTool implements ITool {
    private readonly logger: ILogger;

    constructor(logger: ILogger) {
        this.logger = logger;
    }

    public register(server: McpServer): void {
        server.tool(
            "my-tool",
            "Description of my tool",
            {
                param1: z.string().describe("Parameter description"),
            },
            async ({ param1 }) => {
                this.logger.info(`MyTool called with param1: ${param1}`);
                
                return {
                    content: [
                        {
                            type: "text",
                            text: `Result: ${param1}`,
                        },
                    ],
                };
            },
        );

        this.logger.info("MyTool registered");
    }
}

Adding a New Resource

// resources/custom/my-resource.ts
import { McpServer } from "@modelcontextprotocol/sdk/server/mcp.js";
import { IResource } from "../core/domain/interfaces/resource.interface.js";
import { ILogger } from "../core/domain/interfaces/logger.interface.js";

export class MyResource implements IResource {
    private readonly logger: ILogger;

    constructor(logger: ILogger) {
        this.logger = logger;
    }

    public register(server: McpServer): void {
        server.resource('my-resource', 'my-data.json', async () => {
            this.logger.info('My resource accessed');

            return {
                contents: [
                    {
                        uri: 'my-data.json',
                        text: JSON.stringify({ key: "value" }, null, 2),
                        mimeType: 'application/json',
                    },
                ],
            };
        });

        this.logger.info('My resource registered');
    }
}

Adding a New Prompt

// prompts/custom/my-prompt.ts
import { McpServer } from "@modelcontextprotocol/sdk/server/mcp.js";
import { IPrompt } from "../core/domain/interfaces/prompt.interface.js";
import { ILogger } from "../core/domain/interfaces/logger.interface.js";

export class MyPrompt implements IPrompt {
    private readonly logger: ILogger;

    constructor(logger: ILogger) {
        this.logger = logger;
    }

    public register(server: McpServer): void {
        server.prompt(
            'my-prompt',
            'Description of my prompt',
            () => {
                this.logger.info('My prompt accessed');

                return {
                    messages: [
                        {
                            role: 'user',
                            content: {
                                type: 'text',
                                text: 'Initial user message',
                            },
                        },
                        {
                            role: 'assistant',
                            content: {
                                type: 'text',
                                text: 'Initial assistant response',
                            },
                        },
                    ],
                };
            },
        );

        this.logger.info('My prompt registered');
    }
}

📝 SOLID Principles Applied

Principle	Implementation
Single Responsibility	Each class has one clear responsibility
Open/Closed	System can be extended without modifying existing code
Liskov Substitution	Derived types can be substituted for their base types
Interface Segregation	Specific interfaces for specific clients
Dependency Inversion	High-level modules depend on abstractions

📄 License

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