Skip to content

MCP Model Context Protocol Explained: What It Is & Why You Need It

What is MCP and Why is it Needed? (Model Context Protocol) <br>### The Problem Before MCP <br>When building any AI application (like a travel planner), you often need to integrate various external services such as weather APIs, databases, Google Drive, flight status APIs, and GitHub. Without MCP, a developer must: <br><br>* Write custom code for every single service. Each API has its own unique request/response format and authentication mechanism. <br>* Learn each service's individual documentation. This adds significant development time and complexity. <br>* Rewrite the same logic for different AI frameworks. If you build an app in LangGraph but later want to move it to CrewAI, Open AI Agents SDK, or Agno, you must rewrite all the integration code for that new framework. The app's logic stays the same, but the code implementation changes completely. <br><br>### What is MCP? <br>MCP stands for Model Context Protocol. It is a standard that provides a universal way for AI applications to connect with external data sources and tools. Instead of writing direct code for each service, you connect your AI app to an MCP Server. For a deeper dive into MCP servers and their role in AI development, see our Understanding MCP Servers: A Comprehensive Guide. <br><br>### How MCP Solves the Problem <br>The core idea is abstraction. The MCP server becomes the middle layer that handles all the complexities of interacting with an external tool, be it an API, a database, a local file system, or an internal company system. <br><br>## Core Components of MCP <br>MCP works on a client-server architecture with two main components: <br><br>### 1. MCP Client <br>* The MCP Client is a bridge inside your AI application. <br>* It establishes and manages the connection between your AI app and one or more MCP Servers. <br>* Your AI app must have an MCP Client to use MCP. <br>* A single MCP Client can connect to multiple MCP Servers. <br><br>### 2. MCP Server (and its Types) <br>The MCP Server is where the actual work happens. It connects to the external service and exposes a set of tools that your AI app can use. To learn more about building and mastering MCP servers from scratch, check out Exploring mCP: From Novice to Pro in AI Development. <br><br>There are three types of MCP Servers: <br><br>* Local MCP Server: Runs on the same machine as your AI application. Communication between the client and server happens via Standard I/O (stdin/stdout). <br>* Remote MCP Server: Runs on a cloud or any remote computer. The app connects via the internet, usually using Streamable HTTP. This is ideal for scenarios where you need to access hosted services like a central weather database. <br>* Custom MCP Server: You build this yourself for private or specialized needs (e.g., connecting to your company's internal database). You can host it locally or remotely. <br><br>## How MCP Works: A Step-by-Step Flow <br>Here is the exact process when a user sends a query like "What is the weather in Delhi today?": <br><br>1. User Query: The query goes to the AI application. <br>2. LLM Decision: The application's Large Language Model (LLM) receives the query. If it does not have the answer, it checks a list of available tools. Understanding how LLMs and agents differ is crucial here; read Understanding Generative AI, AI Agents, and Agentic AI: Key Differences Explained. <br>3. Tool Discovery: When the app starts, the MCP Client asks all connected MCP Servers, "What tools do you have?" The servers respond, and the client provides this list to the LLM. For a weather query, the LLM might pick a tool named get_current_weather from the Weather MCP Server. <br>4. Tool Request: The LLM generates a tool request in JSON-RPC format. This is the standard message format for all MCP communication. <br>5. Client-Server Connection: The MCP Client sees the tool request, identifies it belongs to the Weather MCP Server, and establishes the connection. <br>6. Server Execution: The Weather MCP Server executes the request, calls the actual external weather API, and gets the result. <br>7. Response in JSON-RPC: The server formats the raw data (e.g., temperature 38°C, Sunny) back into a JSON-RPC response and sends it to the MCP Client. <br>8. LLM Outputs to User: The client passes this structured data to the LLM, which converts it into a human-readable answer and displays it to the user. For more on turning raw data into effective interactions, see Mastering AI with Context Engineering for Effective Human-AI Collaboration. <br><br>Key takeaway: The LLM decides which tool to use. The MCP Client handles the connection and message transfer. The MCP Server handles the actual execution with the external service. If you're new to AI terminology, check out 20 Key AI Terms Every Engineer Must Know Explained Simply. <br>

Keep this summary

Save it to LunaNotes and it becomes a real note in your library — editable, searchable, and ready to turn into flashcards or a diagram. Free to start.

Save to LunaNotes

Or summarise for another video.

This summary and transcript were automatically generated using AI with the Free YouTube transcript summary tool by LunaNotes.

Related summaries

Understanding MCP Servers: A Comprehensive Guide

Understanding MCP Servers: A Comprehensive Guide

In this video, we explore the concept of Model Context Protocol (MCP) servers, their significance in the AI landscape, and how to create your own MCP server. The discussion covers the functionality of MCP servers, their role in providing context to AI models, and practical examples of implementation.

Exploring mCP: From Novice to Pro in AI Development

Exploring mCP: From Novice to Pro in AI Development

This video takes viewers on a journey from knowing nothing about the Model Context Protocol (mCP) to exploring its features and potential applications. The presenter shares insights on using Klein for seamless mCP server installation, integrating with tools like Superbase, and the challenges faced while navigating the mCP landscape.

20 Key AI Terms Every Engineer Must Know Explained Simply

20 Key AI Terms Every Engineer Must Know Explained Simply

This video breaks down 20 essential AI and large language model (LLM) terms, from tokenization and vectors to RAG and reinforcement learning. Designed for engineers, it clarifies confusing jargon and connects concepts like transformers, fine-tuning, and MCP to real-world applications.

Mastering AI with Context Engineering for Effective Human-AI Collaboration

Mastering AI with Context Engineering for Effective Human-AI Collaboration

Explore practical strategies to harness AI's full potential through context engineering, chain of thought reasoning, role assignment, and reverse prompting. Learn how to transform AI into a reliable teammate for tasks like writing emails, preparing difficult conversations, and enhancing critical thinking.

Claude Full Masterclass 2025: The Complete Guide in Tamil

Claude Full Masterclass 2025: The Complete Guide in Tamil

This is a comprehensive masterclass on Anthropic's Claude AI ecosystem, covering Claude Chat, Claude Code, and the powerful Claude Works. The video explains how to master the entire Claude ecosystem, move from a current stage to 2x or 3x productivity, and use agentic tools without any coding knowledge. It includes a detailed guide on installation, setup, skills, connectors, and plugins, as well as advanced prompt engineering frameworks like CRISP.

Found this summary useful?

Take it with you. One click puts it in your own LunaNotes library.

Save to LunaNotes

Start taking better notes today with LunaNotes