MCP & AI Plugins Tutorials — Model Context Protocol

1 Beginner 10 min read

What is MCP? (And Why Should You Care?)

When I first heard "Model Context Protocol" I thought it was some deeply technical thing I'd never understand. Turns out, the concept is actually pretty simple: MCP is a standard way for AI tools to connect to external data and services. Think of it like USB for AI — a universal plug that lets different tools talk to each other.

Without MCP, AI tools like Claude can only work with what you type into them. With MCP, they can access your files, databases, APIs, and other tools. That's a big deal.

Model Context Protocol architecture overview

MCP connects AI models (clients) to external data and tools (servers) through a standard protocol

The three things MCP provides

Tools

Functions the AI can call — like querying a database, sending an email, or creating a file. The AI decides when to use them based on your request.

Resources

Data the AI can read — like files, documents, or API responses. Think of it as giving the AI access to a library of information.

Prompts

Pre-written templates for specific tasks. Like having ready-made instructions the AI can use for common workflows.

Read the Official MCP Docs Source: modelcontextprotocol.io (Anthropic)

2 Intermediate 20 min read

Build Your First MCP Server

This tutorial from the official MCP docs walks you through building a simple MCP server step by step. The example builds a weather server that gives Claude the ability to check weather alerts and forecasts. Even if you don't care about weather, the pattern is what matters — once you understand it, you can build a server for anything.

You can follow along in Python, TypeScript, Java, C#, or Ruby. I went with TypeScript since that's what I'm learning.

Building an MCP server — define tools, implement handlers, connect to Claude

The basic steps

terminal

              # Install the MCP SDK (pick your language)

              # For Python:

              pip install mcp

              # For TypeScript/Node:

              npm install @modelcontextprotocol/sdk

              # Then follow the tutorial to:

              1. Define your tools (what the AI can do)

              2. Write handler functions (what happens when a tool is called)

              3. Connect to Claude Desktop via config

              4. Test it — ask Claude to use your new tools

You don't need to build from scratch

Check out Tutorial 4 below — there are 500+ pre-built MCP servers for common services. You can use these without writing any code. Building your own is for when you need something custom.

Follow the Build Tutorial Source: MCP Official Docs

3 Beginner Free course

Anthropic's Free MCP Course

If you're more of a structured-learning person (which I am), Anthropic has a free course on their Academy platform that walks through MCP fundamentals. It covers building servers and clients, understanding the three primitives (tools, resources, prompts), and practical integration patterns.

I did this course before trying to build anything, and it saved me a lot of confusion. The course is designed for developers but you don't need to be an expert — basic familiarity with Python or JavaScript is enough.

Anthropic Academy — structured, free learning about MCP from the team that built it

The course covers the "why" behind MCP (not just the "how"), which I found really helpful for understanding when to use it versus when a simple API call would be enough. It also covers security considerations — important stuff if you're connecting AI to real data.

Start the Free MCP Course Source: Anthropic Academy

4 Beginner 5 min to set up

500+ Pre-Built MCP Servers (No Coding Required)

This is the "easy mode" for MCP — and honestly where I'd recommend most beginners start. The official MCP GitHub repo has a directory of 500+ pre-built servers for popular services. GitHub, Google Drive, Slack, Postgres, Puppeteer, file system access — you name it, there's probably a server for it.

You don't need to write any code. Just find the server you want, follow the installation instructions, add a few lines to your config file, and restart Claude. Suddenly Claude can interact with that service.

The official MCP servers directory — browse hundreds of pre-built integrations

How to add a pre-built server to Claude Desktop

claude_desktop_config.json

              // Find the config file:

              // Mac: ~/Library/Application Support/Claude/

              // Windows: %APPDATA%\Claude\

              // Add the server to "mcpServers":

              {

                "mcpServers": {

                  "filesystem": {

                    "command": "npx",

                    "args": [

                      "@modelcontextprotocol/server-filesystem",

                      "/path/to/your/folder"

                    ]

                  }

                }

              }

              // Save the file, restart Claude Desktop

              // Claude can now read/write files in that folder!

Popular servers to try first

Filesystem

Let Claude read and write files on your computer.

GitHub

Give Claude access to your repos, issues, and PRs.

Google Drive

Let Claude search and read your Google Drive files.

Puppeteer

Give Claude a web browser to navigate and screenshot pages.

Browse All MCP Servers Source: MCP Official GitHub

Plugins & MCP