mcp-tutorial

# MCP Tutorial This repository contains the tutorial materials and example code for my **[YouTube course on MCP (Model Context Protocol)](https://www.youtube.com/watch?v=A6chBogLD0k&list=PL_-Bd6t4T02_JnGFZ9iPRa8B4kFeowg9m&index=4&ab_channel=selfdeeplearning)**. The course walks through the evolution of tool use in AI — from **Toolformer** to **MCP** — and then dives into practical examples of building MCP servers, clients, and agents. ## 📺 Course Overview This repository follows along with the YouTube tutorial series: [self deep learning - MCP Course](https://www.youtube.com/watch?v=A6chBogLD0k&list=PL_-Bd6t4T02_JnGFZ9iPRa8B4kFeowg9m&index=4&ab_channel=selfdeeplearning) ### **Chapter 1 — How AI Uses Tools: From Toolformer to MCP** * Who this course is for & prerequisites * What a Tool is & why tools are needed * Classical Tool Use – Toolformer (2023) * Transition to MCP ### **Chapter 2 — MCP Fundamentals** * Core Components — Host, Client, Server * `tools/list` request * `tools/call` request * Protocol Mechanics — JSON-RPC 2.0 * Relationship with Function Calling Features ### **Chapter 3 — MCP Basics in Python** * Setting up the MCP stdio server * Defining MCP tools (e.g., Calculator) * Setting up the MCP stdio client * Calling MCP server tools from the client * Setting up the MCP streamable HTTP server * Setting up the MCP HTTP client ### **Chapter 4 — Connecting to Existing MCP Servers** * Connecting to built-in servers such as **Filesystem**, **Git**, and **Tavily** ### **Chapter 5 — MCP + LLM Integration** * How to connect MCP to an LLM (e.g., OpenAI API, LangChain/LangGraph) * Security risks and safe design considerations ### **Chapter 6 — Future Outlook & Closing** * What we didn’t cover and why * Future outlook for MCP and related approaches ## 📂 Directory Structure ``` . ├── servers/ # MCP servers (stdio, HTTP, etc.) ├── clients/ # MCP clients (stdio, HTTP, etc.) ├── agents/ # Chatbot agents using OpenAI API or LangChain └── assets/ # Slides & YouTube scripts ``` ## 🚀 Getting Started ### 1. Install `uv` Follow the instructions here: [uv installation guide](https://docs.astral.sh/uv/getting-started/installation/). For example, on macOS/Linux: ```bash curl -LsSf https://astral.sh/uv/install.sh | sh ``` On Windows (PowerShell): ```powershell irm https://astral.sh/uv/install.ps1 | iex ``` ### 2. Create Virtual Environment ```bash uv venv source .venv/bin/activate # On macOS/Linux .venv\Scripts\activate # On Windows ``` ### 3. Install Dependencies ```bash uv sync ``` ### 4. Run an Example ```bash uv run clients/stdio_client.py ``` ### 5. Environment Variables Some examples(**tavily_client**, **agents**) in this project require API keys for external services (e.g., LLMs, search tools). 1. Copy `.env.example` to `.env`: ``` cp .env.example .env ``` 2. Edit `.env` and replace with your own keys: ``` TAVILY_API_KEY=tvly-xxxxxx OPENAI_API_KEY=sk-xxxxxx ``` You can get your keys here: 👉 Tavily API Key → https://app.tavily.com/ 👉 OpenAI API Key → https://platform.openai.com/ ## 🎯 Learning Goals By the end of this tutorial project, you will: * Understand what MCP is and why it matters. * Learn how to build your own **MCP servers and clients**. * Connect existing servers (Filesystem, Git, Tavily). * Integrate MCP with LLMs safely. ## 📌 Notes * This repo is designed to be **educational**, not production-ready. * The `assets/` folder contains slides and scripts used in the videos.