MCP Server Using NestJS

The MCP (Model Context Protocol) Server built using NestJS is an extensible, modular framework designed to enable interaction between AI models (like Claude, GPT, etc.) or external clients with specialized tools through HTTP requests. With dynamic tool registration and robust validation mechanisms, it seamlessly integrates into the AI workflow ecosystem.

Overview: What is MCP Server Using NestJS?

The MCP server acting as a mediator uses NestJS, a progressive Node.js framework for building efficient and scalable applications. This project showcases how AI applications can leverage MCP to connect with backend tools and services through standardized protocols. It serves as a base for developers integrating LLMs (like Claude, GPT) into their workflows, ensuring seamless interaction and data handling.

Core Features & MCP Capabilities

The server is built around the core concept of dynamic tool registration and execution via HTTP endpoints. Each tool, such as calculator, temperature, filesystem, code-analyzer-local, is registered with a name, validation schema using Zod, and a handler function. This modular architecture allows for easy expansion and customization, making it highly adaptable to different AI application needs.

Tool Details:

• calculator: Supports basic mathematical operations (addition, subtraction, multiplication, division).

  • Example Request:

    {
      "toolName": "calculator",
      "params": {
        "a": 10,
        "b": 2,
        "operation": "multiply"
      }
    }
    

• temperature: Provides a fictional temperature based on a prompt.

  • Example Request:

    {
      "toolName": "temperature",
      "params": {
        "prompt": "Temperatura en Madrid"
      }
    }
    

• filesystem: Handles file operations like reading, writing, and deletion.

  • Example Request:

    {
      "toolName": "filesystem",
      "params": {
        "path": "src/main.ts",
        "action": "read"
      }
    }
    

• code-analyzer-local: Generates descriptive .md files by analyzing code directories.

  • Example Request:

    {
      "toolName": "code-analyzer-local",
      "params": {
        "directory": "mcp-server/src"
      }
    }
    

MCP Architecture & Protocol Implementation

The architecture of the MCP server is based on modular design, ensuring that tools can be dynamically registered and utilized. Zod plays a crucial role in validating both user inputs and AI model outputs to ensure consistent data flow.

Zod Validation Example:

import { z } from "zod";

// Define expected schema
const TaskSchema = z.object({
  title: z.string(),
  completed: z.boolean(),
});

// Validate LLM output
try {
  const parsed = TaskSchema.parse(JSON.parse(llmOutput));
  console.log("✅ Valid data:", parsed);
} catch (err) {
  console.error("❌ Invalid data:", err);
}

MCP Protocol Flow Diagram

The MCP protocol flow can be visualized as:

graph TD
    A[AI Application] -->|MCP Client| B[MCP Server]
    B --> C[MCP Client]
    C --> D[Data Source/Tool]
    style A fill:#e1f5fe
    style C fill:#f3e5f5
    style D fill:#e8f5e8

Getting Started with Installation

To install the necessary dependencies, navigate to your project directory and run:

npm install zod

Once installed, you can configure the server using .env files:

PORT=3000
ANALYSIS_DIR=./src/docs

Key Use Cases in AI Workflows

Real-world use cases include:

1. Financial Analysis Tool Integration:

   • An LLM model generates financial forecasts based on historical data. The MCP server integrates this forecast with a proprietary financial analysis tool, allowing for real-time interaction and enhanced decision-making.

2. Code Optimization Pipeline:

   • Developers can integrate the code-analyzer-local tool to automatically generate documentation from source code repositories. This streamlines the documentation process and ensures consistency in project documentation.

Integration with MCP Clients

The server supports a range of MCP clients, including:

• Claude Desktop: Fully compatible
• Continue: Fully compatible
• Cursor: Tool-only support

Performance & Compatibility Matrix

Compatibility matrix for MCP clients:

MCP Client	Resources	Tools	Prompts	Status
Claude Desktop	✅	✅	✅	Full Support
Continue	✅	✅	✅	Full Support
Cursor	❌	✅	❌	Tools Only

Advanced Configuration & Security

Configuring the server involves setting up environment variables and specifying server commands. Here's a sample configuration:

{
  "mcpServers": {
    "[server-name]": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-[name]"],
      "env": {
        "API_KEY": "your-api-key"
      }
    }
  }
}

Frequently Asked Questions (FAQ)

1. How does the MCP protocol ensure secure communication between AI applications and tools?
2. What are the benefits of using Zod for validation in the context of AI model outputs?
3. Can multiple LLM models be integrated into a single MCP server, and if so, what are the best practices?
4. How does the MCP server enable real-time interaction between AI applications and backend tools?
5. What steps should I take to ensure compatibility with different MCP clients?

Development & Contribution Guidelines

Contributions can be made by cloning the repository, setting up the environment, and submitting pull requests for new features or fixes.

MCP Ecosystem & Resources

For more resources, developers are encouraged to explore official documentation, courses, video tutorials, and a job board dedicated to AI applications and MCP integrations.

This comprehensive documentation positions the MCP server as a robust solution for integrating AI applications with backend tools, enhancing their capabilities through standardized protocols.