Getting started with Model Context Protocol (MCP)

Model Context Protocol (MCP) is rapidly gaining traction in the AI community and for good reason. Developed by Antropic and released as open-source in November 2024, MCP is emerging as a compelling advancement in how AI systems interact with external data and tools.

What is Model Context Protocol (MCP)?

Here is the definition by the Antropic
MCP is an open protocol that standardizes how applications provide context to LLMs. Think of MCP like a USB-C port for AI applications. Just as USB-C provides a standardized way to connect your devices to various peripherals and accessories, MCP provides a standardized way to connect AI models to different data sources and tools.
What this means, the Model Context Protocol (MCP) let AI applications to connect different tools and data sources in a plug and play way. Think of it as the universal adapter for feeding external context into LLMs. On the other hand, AI agents are built to act they’re autonomous or semi-autonomous systems designed to pursue goals, complete tasks, and use tools along the way. While agents focus on reasoning and execution, MCP focuses on standardising how information like tools, databases, or user data is packaged and delivered to the LLMs.

While AI agents are autonomous or semi-autonomous systems that act and plan using LLMs, the Model Context Protocol (MCP) is the infrastructure layer that standardises how those models receive context. If AI agents are the drivers, MCP is the well paved highway.

MCP vs AI Agents

Model Context Protocol (MCP) Architecture

MCP follows a client-server architecture (Like microservices expose APIs to interact with other services, MCP provides a standardised interface for LLMs to access tools, data, and memory). Below is the MCP architecture by Antropic

Hosts are LLM applications (like Claude Desktop or IDEs) that initiate connections
Clients maintain 1:1 connections with servers, inside the host application
Servers provide context, tools, and prompts to clients

MCP Architecture

Fun Time: Lets try Model Context Protocol (MCP)

⚠️ Disclaimer on MCP Server Usage Important: MCP is an open protocol, which means anyone can develop and deploy an MCP compatible server. While this encourages innovation and flexibility, it also introduces potential risks.

If you’re running MCP servers locally or connecting to external ones, be cautious and only use servers that are validated, trusted, or verified by reputable sources. Malicious actors could build MCP servers that expose sensitive data, execute harmful code, or behave unpredictably.

Always audit the server’s source code when possible

1. Find a Client Application

You can find all the MCP supported client applications here https://modelcontextprotocol.io/clients and I will be using Claude Desktop App and I will install this in my Local machine.
You must sign up to use the application and once its done you can open the Claude Desktop App and proceed to next step.

Claude Desktop App

2. Find a MCP Server

Now you need to find a MCP supported servers, which are validated and trusted. You can find them here https://modelcontextprotocol.io/examples
I will use Reference/Official filesystem server and Community Airbnb server.

Note: Community servers are untested and should be used at your own risk. They are not affiliated with or endorsed by Anthropic.

3. Add a MCP Server to Claude Desktop App

Once you are confident about the Server, you can add the capabilities of these servers into Claude Desktop App

1 - Claude -> Settings

MCP Server Configuration Step 1

2 - Developer -> Edit Config

MCP Server Configuration Step 2

3 - Open claude_desktop_config.json file in a text editor

MCP Server Configuration Step 3

4 - Open filesystem and copy the content

MCP Server Configuration Step 4

Change username to your username

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
{
  "mcpServers": {
    "filesystem": {
      "command": "npx",
      "args": [
        "-y",
        "@modelcontextprotocol/server-filesystem",
        "/Users/username/Desktop",
        "/path/to/other/allowed/dir"
      ]
    }
  }
}