MCP Overview

MCP is a protocol for building agents that can interact with external tools. It is a way to connect agents to the outside world.

Making LLMs Truly Useful: The Role of Tool Use and Anthropic’s MCP

Large Language Models (LLMs) are powerful, but they have a major limitation:
LLMs cannot access real-time or live data out of the box.
To overcome this, LLMs need the ability to call external tools or functions—essentially, to perform actions or retrieve information beyond their training data.

The Problem with Traditional Tool Use

Traditionally, if you want an LLM to do something it wasn’t trained on—like deleting files in Google Drive or renaming documents—you have to define a function for each specific task. For example:

def delete_file_in_drive(file_id):
    # code to delete file
    pass

def rename_file_in_drive(file_id, new_name):
    # code to rename file
    pass

This approach quickly becomes unwieldy. As the number of tasks grows, so does the complexity of your codebase, making it hard to scale and maintain.

Enter Anthropic’s MCP

To address these challenges, Anthropic introduced the MCP (Modular Command Protocol).
MCP is a set of rules that every tool or service must follow to provide context and capabilities to LLMs in a scalable way.

MCP’s Three-Tier Architecture

MCP is designed with a clear, three-tier architecture:

• MCP Host:
  The platform or environment that acts as a bridge between the user, the LLM, and the MCP Server.
  Examples: Cursor, Windsurf

• MCP Client:
  Protocol clients that maintain a direct connection with the servers.
  Example: Internal code in Cursor IDE that communicates with the MCP Server

• MCP Server:
  The backend server that understands the MCP protocol and provides the actual services.

How MCP Communication Works

Here’s a high-level overview of the MCP communication workflow:

Why MCP Matters

All MCP servers, regardless of their underlying system, speak the same protocol.
When a service updates its MCP server with new methods, those methods automatically become available to you—no extra code required. This means you can scale your LLM’s capabilities without ballooning your codebase.