MCP Server — The USB Port for AI Tools

In the age of AI, the term MCP Server keeps popping up everywhere.
So what exactly is it?
Is it a framework? An architecture pattern?
Let’s break it down in the simplest way.

---

What Is MCP?

MCP (Model Context Protocol) is an open standard from OpenAI that allows:

• AI clients (ChatGPT, VS Code, CLIs)
• To access external tools and data
• Using standardized MCP-compatible servers

Think of MCP like USB for AI:
If a tool is MCP compatible, any AI client can plug into it.

---

Why MCP Exists?

MCP gives us:

• Consistent communication format
• Secure interactions
• Plug-and-play tools
• Cross-LLM compatibility

One server → works with multiple AI platforms.

---

Core Components of MCP

1. MCP Client

Where the AI runs:
ChatGPT, VS Code, terminal tools, or your own software.

2. MCP Server

Provides access to:

• APIs
• Files
• Databases
• Functions, scripts, and workflows
  All wrapped in MCP’s common format.

3. Protocol Messages

The “language” between client and server:

• list tools
• fetch resource
• call tool
• stream events

---

Practical Things an MCP Server Can Do

Examples of capabilities you can expose:

🗄️ Read/write files
🔎 Search organization documents
📬 Fetch emails
💳 Call internal microservices
⚙️ Run command-line scripts
📊 Query SQL/NoSQL databases
🤝 Integrate Jira, GitHub, Slack, Notion, etc.

This changes AI from answering questions → taking actions.

---

Why Teams and Developers Should Care?

• Faster integration across AI tools
• Shared, reusable tool modules
• Security control: expose only what you allow
• Zero vendor lock-in
• Future-proof: works with multiple AI systems

---

Build the Simplest MCP Server in Python

Install the SDK:

pip install mcp

Create simple_mcp_server.py:

#!/usr/bin/env python3
from mcp.server import Server
from mcp.types import TextContent, Tool

server = Server(name="simple-mcp")

@server.resource("hello.txt")
async def hello_resource():
    return TextContent("Hello from MCP resource!")

@server.tool(
    Tool(
        name="echo",
        description="Return back whatever message you send",
        input_schema={"type": "object","properties":{"message":{"type":"string"}},"required":["message"]}
    )
)
async def echo_tool(message: str):
    return TextContent(f"You said: {message}")

@server.tool(
    Tool(
        name="add",
        description="Add two numbers",
        input_schema={
            "type": "object",
            "properties": {"a": {"type": "number"}, "b": {"type": "number"}},
            "required": ["a", "b"],
        },
    )
)
async def add_tool(a: float, b: float):
    return TextContent(f"{a} + {b} = {a + b}")

if __name__ == "__main__":
    server.run_stdio()

---

How this Server Works?

Resources

• hello.txt

  • Read-only
  • Discoverable by AI

Tools

• echo → returns your message
• add → simple math Both enforce a JSON schema for safety.

Minimal surface

The AI only calls what you expose.

---

Running the Server

python simple_mcp_server.py

MCP servers do not expose HTTP endpoints. Instead, they communicate using:

STDIN  ←→  STDOUT

ChatGPT or VS Code launches your script and exchanges JSON messages over pipes.

---

Connect Your MCP Server to VS Code

✔ Prerequisites

• VS Code
• Python 3.10+
• MCP SDK installed

Step 1: Install Extension

Open VS Code → Extensions → Search Model Context Protocol → Install

Step 2: Create Folder

~/mcp-test/

Open it in VS Code

Step 3: Add Settings

Create:

.vscode/settings.json

Add:

{
  "mcpServers": {
    "simple-mcp": {
      "command": "python",
      "args": ["simple_mcp_server.py"]
    }
  }
}

Step 4: Load Server

Open Command Palette → MCP: Reload Servers

VS Code should detect:

• hello.txt
• echo
• add

Step 5: Try a Call

Send:

{
  "tool": "echo",
  "params": { "message": "Hello VS Code!" }
}

You’ll see:

"You said: Hello VS Code!"

---

Why MCP Is a Game Changer?

MCP is the foundation for AI-native engineering.

Soon:

• Companies will run many MCP servers
• AI systems will orchestrate them automatically
• Developers won’t bolt AI onto APIs — they’ll build tools for AI

---

Final Thoughts

MCP transforms AI from a passive assistant into a powerful system operator. With just a few lines of code, your AI can read files, trigger scripts, and talk to real systems.

Happy building — and welcome to the future of AI tooling 🚀