A Beginner’s Guide to the Model Context Protocol (MCP):

source: https://x.com/Meer_AIIT/status/2015091824851836979

This isn’t a clickbait article that skims the surface. I’ve personally learned MCP from the ground up.

Worked through multiple courses.

Studied the official documentation.

Built servers myself.

I’ll link to all my sources so you can dive deeper but my goal here is to save you time by distilling everything into one comprehensive guide.

By the end of this article you’ll understand the complete fundamentals.

• What MCP servers actually are.
• How to use them.
• Why they matter.

And how to build your own… both with code and without it.

What is MCP?

The Model Context Protocol (MCP) is an open standard that enables AI applications to connect with external tools and data sources in a standardized way.

Think of it as a universal language that lets AI models talk to the outside world.

Before MCP if you wanted Claude or another AI to access your Google Drive you’d need custom code.

Want it to also read from Slack?

More custom code.

Each integration required its own unique solution.

MCP changes this by providing a single consistent protocol.

Build once. Connect to everything.

image source: MCP: Build Rich-Context AI Apps with Anthropic

The USB-C Analogy:

Here’s a helpful way to think about it.

Trending

Walker & Dunlop Arranges $104.5 Million Loan for Ritz-Carlton Savannah Construction in Historic District

MCP is like USB-C for AI applications.

Remember when every device had its own charger?

Phones. Tablets. Laptops. All different cables.

USB-C standardized this mess.

Now one cable works for almost everything.

MCP does the same thing for AI integrations.

Instead of building custom connectors for every tool or database… developers implement the MCP standard once.

Their AI can then connect to any MCP-compatible service through a single unified interface.

file systems databases APIs enterprise tools

All through one protocol.

AI generated image

Who Introduced MCP:?

MCP was announced by Anthropic in November 2024 as an open standard for connecting AI assistants to data systems.

Content repositories. Business management tools. Development environments.

The protocol emerged from a practical problem Anthropic’s team kept encountering.

The “N×M integration problem.”

As the number of AI applications (N) grows and the variety of tools and data sources (M) increases… the complexity of creating custom integrations for each combination becomes unmanageable.

The protocol was released with SDKs in Python, TypeScript, C# and Java.

Anthropic open-sourced everything from day one.

Since launching in November 2024 adoption has been rapid.

The community has built thousands of MCP servers.

SDKs are available for all major programming languages.

The industry has adopted MCP as the de-facto standard for connecting agents to tools and data.

Following its announcement the protocol was adopted by major AI providers including OpenAI and Google DeepMind.

Core Concepts & Architecture

MCP uses a straightforward client-server architecture.

image source: MCP: Build Rich-Context AI Apps with Anthropic

Understanding this structure is key to working with the protocol effectively.

The Three Main Components:

Hosts: are the AI applications you interact with.

Claude Desktop, An IDE with AI features, A custom chatbot.

The host manages the overall experience and coordinates between the AI model and external services.

Clients: live inside hosts and handle communication with MCP servers.

Think of them as translators that speak the MCP language on behalf of the host application.

Servers: are lightweight programs that expose specific capabilities.

A server might provide access to your file system.

Connect to a database.

Or integrate with a business tool like Slack or GitHub.

The Three Core Primitives:

MCP servers can expose three types of capabilities.

Tools: are functions the AI can call to perform actions.

Searching files. Querying a database. Sending a message. Making an API request.

When Claude needs to “do something” it uses a tool.

Resources: provide data and content the AI can read as context.

Documents. Database records. Configuration files.

Resources give the AI information to work with.

Prompts: are reusable templates and workflows.

They help standardize how the AI approaches certain tasks ensuring consistent and high-quality outputs.

How Communication Works

When you ask Claude to “find all my project files modified this week” here’s what happens behind the scenes:

1. Claude connects to your file system MCP server via the host and client
2. The client requests the available tools from the server
3. Claude identifies the right tool and calls it with appropriate parameters
4. The server executes the request and returns results
5. Claude processes the results and responds to you

All of this happens through a standardized message format.

Typically using JSON over stdio for local servers or HTTP for remote servers.

Why MCP Matters

Problems It Solves

Eliminates integration fragmentation.

Without MCP every combination of AI app and external tool requires custom code.

For 10 AI apps connecting to 20 tools that’s potentially 200 unique integrations.

MCP reduces this to 10 + 20 = 30 implementations.

Enables true AI agents.

AI becomes dramatically more useful when it can take action… not just discuss things.

MCP provides the secure standardized way for AI to interact with real-world systems.

Maintains security and control.

Every action requires your approval.

The host controls which servers connect.

You approve each tool invocation.

Your data stays protected while still being accessible to AI.

Reduces vendor lock-in.

Because MCP is an open standard your integrations work across different AI providers.

Switch from one model to another without rebuilding your tool connections.

Real Benefits

For developers MCP means building an integration once and having it work everywhere.

For organizations it means AI assistants that can actually access internal systems and data.

For users it means AI that’s genuinely helpful.

Not just conversational but capable.

Getting Started

No-Code Approach: Desktop Extensions

The easiest way to experience MCP is through Claude Desktop’s extension system.

Desktop extensions provide a streamlined way to install and manage local MCP servers through single-click installable packages.

Instead of manually configuring JSON files and managing dependencies… you can now install local MCP servers as easily as browser extensions.

Here’s how to get started:

1. Download and install Claude Desktop from claude.ai
2. Open Claude Desktop and go to Settings > Extensions
3. Browse the extension directory for Anthropic-reviewed tools
4. Click “Install” on any extension that interests you
5. Configure any required settings like API keys through the interface

That’s it.

The extension handles all the technical setup automatically.

Low-Code Approach: Pre-Built Servers

If you’re comfortable with a terminal you can install pre-built MCP servers manually.

This gives you more options than the extension directory.

First ensure Node.js is installed on your system.

Open a terminal and run:

bash

node -v

If you see a version number you’re ready.

If not download Node.js from nodejs. org.

Next configure Claude Desktop to use a server.

Open the Claude Desktop configuration file:

macOS: ~/Library/Application Support/Claude/claude_desktop_config.jsonWindows: %APPDATA%\Claude\claude_desktop_config.json

Add a server configuration.

Here’s an example for the filesystem server:

json

{
  "mcpServers": {
    "filesystem": {
      "command": "npx",
      "args": [
        "-y",
        "@modelcontextprotocol/server-filesystem",
        "/Users/yourname/Desktop",
        "/Users/yourname/Documents"
      ]
    }
  }
}

Replace the paths with directories you want Claude to access.

Save the file and restart Claude Desktop.

Developer Approach: Building Your Own

For full control you can build custom MCP servers using the official SDKs.

This requires programming knowledge but offers unlimited flexibility.

We’ll cover a basic example in the next section.

Simple Implementation Example

Let’s build a minimal MCP server in Python.

This server will expose one tool: a calculator that evaluates mathematical expressions.

Setup

First create a project directory and install the MCP SDK:

bash

# Create and enter project directory
mkdir my-mcp-server
cd my-mcp-server

# Create virtual environment (recommended)
python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

# Install the MCP SDK
pip install mcp

The Server Code

Create a file called calculator_server.py:

python

from mcp.server.fastmcp import FastMCP

# Initialize the MCP server with a name
mcp = FastMCP("Calculator Server")

# Define a tool using the @mcp.tool() decorator
# The docstring becomes the tool's description for the AI
@mcp.tool()
def calculate(expression: str) -> str:
    """
    Evaluate a mathematical expression and return the result.
    
    Args:
        expression: A math expression like "2 + 2" or "10 * 5 / 2"
    
    Returns:
        The calculated result as a string
    """
    try:
        # eval() is used here for simplicity - in production,
        # use a safer math parser like numexpr or simpleeval
        result = eval(expression)
        return f"Result: {result}"
    except Exception as e:
        return f"Error: {str(e)}"

# Run the server using stdio transport (for local use)
if __name__ == "__main__":
    mcp.run(transport="stdio")

What’s Happening Here

The FastMCP class handles all the protocol details for you.

When you decorate a function with

@mcp

.tool() it automatically becomes available as an MCP tool.

The function’s type hints (expression: str) and docstring are parsed to create the tool’s schema.

This tells AI models what parameters the tool expects and what it does.

Testing Your Server

You can test your server using the MCP CLI:

bash

mcp dev calculator_server.py

This opens a local dashboard where you can interact with your tools directly.

Connecting to Claude Desktop

Add your server to Claude Desktop’s configuration:

json

{
  "mcpServers": {
    "calculator": {
      "command": "python",
      "args": ["/full/path/to/calculator_server.py"]
    }
  }
}

Restart Claude Desktop and you can now ask Claude to calculate things.

It will use your server to perform the actual math.

Common Use Cases

MCP opens up a wide range of practical applications.

File and document management

Let AI organize your downloads. Search through documents. Create and edit files on your system.

Database access

Connect AI to PostgreSQL, MySQL or other databases.

Ask questions about your data in plain English and get actual query results.

Development workflows

Integrate with GitHub for managing issues and pull requests.

Connect to your IDE for AI-assisted coding.

Business tool integration

Connect to Slack. Google Drive. Notion. Salesforce.

Build AI assistants that understand your organization’s context.

Custom automation

Build servers that wrap your own APIs and internal tools.

Create AI-powered workflows specific to your needs.

Research and analysis

Give AI access to data sources and web scraping tools.

Let it gather and process information for you.

Key Takeaways

MCP represents a fundamental shift in how AI applications connect to the world.

Instead of isolated chatbots we now have AI that can actually participate in our digital workflows.

The protocol is open. Standardized. Already widely adopted.

Whether you’re a casual user installing extensions or a developer building custom integrations… MCP provides a path forward.

Security remains in your hands.

Every connection is controlled.

Every action requires approval.

Your data stays protected. Here are links of each sources from I’ve learned entire MCP Servers: 1.

2.

3.

If you want to stay updated with the latest AI news and tools and research paper breakdowns… subscribe to my free newsletter: