How to Create an MCP Server – Step-by-Step Tutorial

​Modern AI applications are no longer just chatbots. They need tools, memory, APIs, and structured context. That’s exactly where MCP (Model Context Protocol) comes in.

In this tutorial, you’ll learn what an MCP server is, why it matters, and how to create an mcp server from scratch using Node.js or Python.

What Is an MCP Server?

An MCP server is a backend service that exposes tools, resources, and prompts in a standardized way so that AI models (like Claude or other MCP-compatible LLMs) can interact with your system safely and deterministically.

Instead of giving an LLM direct access to your database or APIs, you expose controlled capabilities through MCP.

MCP Solves These Problems

• ❌ Unstructured prompt hacking
• ❌ Unsafe direct API access
• ❌ Tight coupling between LLMs and backend
• ❌ Hard-to-scale agent tooling

✅ MCP provides:

• Tool discovery
• Typed inputs/outputs
• Secure boundaries
• Plug-and-play AI integrations

---

MCP Architecture Overview

Flow:

1. LLM connects to MCP client
2. MCP client connects to your MCP server
3. Server exposes tools/resources
4. LLM calls tools with structured input
5. Server responds with typed output

---

Prerequisites

Before starting, you should know:

• Basic JavaScript or Python
• REST APIs
• JSON & async programming

Tools Needed

• Node.js 18+ or Python 3.10+
• npm / pip
• Terminal

---

Option 1: Create an MCP Server Using Node.js

Step 1: Initialize the Project

mkdir my-mcp-server
cd my-mcp-server
npm init -y

Step 2: Install MCP SDK

npm install @modelcontextprotocol/sdk zod

---

Step 3: Create the Server File

Create server.js:

import { Server } from "@modelcontextprotocol/sdk/server/index.js";
import { StdioServerTransport } from "@modelcontextprotocol/sdk/server/stdio.js";
import { z } from "zod";

---

Step 4: Define the MCP Server

const server = new Server(
  {
    name: "my-first-mcp-server",
    version: "1.0.0",
  },
  {
    capabilities: {
      tools: {},
    },
  }
);

---

Step 5: Add a Tool

Example: Text to Uppercase Tool

server.tool(
  "toUpperCase",
  {
    text: z.string(),
  },
  async ({ text }) => {
    return {
      content: [
        {
          type: "text",
          text: text.toUpperCase(),
        },
      ],
    };
  }
);

---

Step 6: Start the MCP Server

const transport = new StdioServerTransport();
await server.connect(transport);

Run it:

node server.js

🎉 Your MCP server is now live.

---

Option 2: Create an MCP Server Using Python

Step 1: Install Dependencies

pip install mcp pydantic

---

Step 2: Create the Server

Create server.py:

from mcp.server import Server
from pydantic import BaseModel

---

Step 3: Define Tool Input Schema

class UpperInput(BaseModel):
    text: str

---

Step 4: Register Tool

server = Server("python-mcp-server")

@server.tool()
def to_upper_case(input: UpperInput):
    return input.text.upper()

---

Step 5: Run Server

server.run()

---

Exposing Resources (Files, Data, APIs)

You can expose read-only resources such as logs, config files, or database views.

Example:

server.resource(
  "config",
  "config://app",
  async () => ({
    contents: [
      {
        uri: "config://app",
        text: JSON.stringify(process.env, null, 2),
      },
    ],
  })
);

---

MCP Server vs REST API

Feature	REST API	MCP Server
AI-safe	❌	✅
Tool discovery	❌	✅
Typed schema	❌	✅
Prompt injection safe	❌	✅
LLM-native	❌	✅

---

Best Practices for MCP Servers

✅ Keep Tools Small

Each tool should do one thing well

✅ Validate All Inputs

Always use schemas (zod / pydantic)

✅ Never Expose Secrets

Use server-side authentication & environment variables

✅ Log Tool Calls

Critical for debugging AI behavior

---

Real-World Use Cases

• AI agents for DevOps
• Database query assistants
• Code analysis tools
• Internal company copilots
• Secure enterprise chatbots

---

Common Mistakes to Avoid

❌ Putting business logic in prompts
❌ Returning unstructured text
❌ Overloading one tool with many actions
❌ Allowing unrestricted file access

---

Final Thoughts

MCP is quickly becoming the standard way to connect AI models with real systems. By building an MCP server, you future-proof your AI applications and gain security, structure, and scalability.

If you already build APIs, learning MCP will feel natural—but far more powerful.