MCP 🧠 Simplified: A Beginner's Guide

---

📚 Topic Covered In This Article

✨ What is MCP ?

✨ MCP Architecture .

✨ What is Tools in MCP ?

✨ What is Resources in MCP ?

✨ What is Prompts in MCP ?

---

In this blog series , we’ll explore the Model Context Protocol ( MCP ) - a powerfull standard that connects LLMs with tools, APIs and external data sources .

This series is divided into 3 parts :

1. 📘 Part 1 : Understanding MCP

2. 🛠️ Part 2 : Building an MCP Server .

3. 🤖 Part 3 : Integrating MCP Server with Claude Desktop and with our own AI Agents

🤔 What is MCP ?

Let’s start by breaking MCP (Model Context Protocol)

In simple term , MCP is a standardized way ( or protocol ) to provide information and context to Large Language Models (LLMs) .

It allows LLMs to interact with external tools , APIs and data sources in a structured way . instead of manually injecting data into prompts or building custom logic for each integration, MCP enables a clean , client - server architecture where tools can easily send context to model and receive intelligent responses.

Think of MCP as a universal bridge 🌉, enabling AI models to access the information they need to perform task effectively and enabling developers to build more powerful and scalable AI Applications .

---

🧠 Example : AI Assistant + MCP in Action

Let’s say you’re building an AI Assistant to help users write emails based on recent meetings . your system supports MCP , and the AI Assistant is connected to tools like :

• 📆 Calendar API

• 📝 Meeting Notes Service (Internal Database that consist of meeting Notes / transcript )

• 📧 Email Client

All these tools expose data via standardized APIs call and MCP acts as a bridge to access them based on context .

🗣️ When the user Asks :

“Write a follow-up email for yesterday’s client meeting with John”

Without MCP, the AI Assistant doesn’t know:

• What meeting the user is referring to

• Who attended

• What was discussed

So, the assistant lacks the necessary context / information to generate a meaningful follow-up email.

✅ Thanks to MCP...

The AI Assistant automatically receives relevant context like :

• Meeting Title : “Q3 Marketing Strategy” .

• Attendees : You and John .

• Key Points : Launch new campaigns in July , finalize budget next week .

• Meeting Transcript: Full conversation notes from the meeting .

Using this structured context AI instantly generates a follow-up email that says:

Hi John

Great speaking with you yesterday about the Q3 marketing strategy . As discussed we’ll move forward with launching the campaign in July…

🔎 What is Happening Under the Hood ?

• Tools like Calendar and notes service are just APIs or databases that exposes relevant data

• MCP standardized how any AI application can access and uses this context .

• No need for the user to copy-paste or retype information .

• MCP provide a clean contract( like a plugin interface) for tools to expose capabilities .

---

🧱 Architecture of MCP :

https://modelcontextprotocol.io/introduction

At its core, MCP (Model Context Protocol) follows client-server architecture , where a host application ( like Claude Desktop or an AI agent ) connects to one or more MCP servers to fetch or send contextual data from various sources - both local and remote.

🔍 Key Components

• 🧠 MCP Hosts (Clients)

Applications like Claude Desktop , IDEs or custom AI tools that want to access structured context ( files , databases , APIs ) using MCP protocol .

• 🔗 MCP Clients

The MCP Client is what allows your AI application to talk to MCP servers . It maintains a 1:1 connections with an MCP server

• 🖥️ MCP Servers

An MCP Server is like an adapter for application or service . It knows how to take a request from an AI — Like : Get me today’s sales report — and translate it into the right commands or API calls that the underlying tool understands .

• 📂 Local Data Sources

Your computer internal resources — such as files , folders databases or running services that an MCP server can securely connect to and expose to the AI agent .

• 🌐 Remote Services

External web services or APIs (like Weather APIs etc ) that MCP servers can connect to and present as usable context to the AI .

🧩 Core Concepts of MCP

To fully understand how MCP works , lets break down a few important terms :

• 🔧 Tools — Let the AI Take Action

Tools allows your AI agent to do things - like call APIs , trigger scripts or update systems — via your MCP server .

You can think of tools as executable functions exposed by the server . Once a tool is registered , The AI can call by making requests .

📌 Tools are for action — they let the AI interact with external systems in real time.

• 🗂️ Resources — Provide the AI with Data

Resources are pieces of data your MCP server exposes to the AI . These could be files , database entries , logs , images or even live systems data .

Examples :

1. PDFs, documents and code files

2. Database query results

3. Log files or screenshots

📌 Resources are for information — they give the AI something to read or analyze.

• ✏️ Prompts — Reusable Interactions

Prompts are predefined templates or workflows that make AI interactions faster and more consistent . Think of them like reusable commands or prefilled instructions for the LLM .

📌 Prompts are for guidance — they help the user trigger common or complex LLM interactions with minimal effort.

🚀 Wrapping Up

In this first part of the series , we explored the Model Context Protocol (MCP) — a powerful standard that connects Large Language Model (LLMs) to tools , data sources and services in a clean , structured way .

Here’s what we covers :

• What MCP is and the real - world pain point it solves

• A practical example showing how AI tools can work together with shared context

• The core architecture of MCP

• Key MCP concepts like Tools , Resources and Prompts

🔜 Coming Up in Part 2

In the next post we’ll get hands-on and :

• 🔧 Build a working MCP server from scratch .

• 📂 Expose tools and resources to the AI .

• 📡 Test real-time context exchange between client and server .

💬 Got feedback, questions, or something you'd love to see covered next? Drop it in the comments

Thanks for reading — 🙏🤝 👋 see you in Part 2!