这是一个面向初学者的开源 MCP 课程，由 Microsoft 维护，通过 .NET、Java、TypeScript、JavaScript、Rust 和 Python 的实战代码，讲解模型上下文协议的基本概念与实现方式。内容覆盖 MCP 服务器与客户端、工具集成、安全实践、服务编排和部署，并配有文档、规范链接与多语言翻译。它更像一套循序渐进的学习路径，帮助开发者理解 AI 应用如何以标准化方式连接外部工具和数据，适合边看边练，也可作为团队内部培训材料。
git clone https://github.com/microsoft/mcp-for-beginners.gitArabic | Bengali | Bulgarian | Burmese (Myanmar) | Chinese (Simplified) | Chinese (Traditional, Hong Kong) | Chinese (Traditional, Macau) | Chinese (Traditional, Taiwan) | Croatian | Czech | Danish | Dutch | Estonian | Finnish | French | German | Greek | Hebrew | Hindi | Hungarian | Indonesian | Italian | Japanese | Kannada | Khmer | Korean | Lithuanian | Malay | Malayalam | Marathi | Nepali | Nigerian Pidgin | Norwegian | Persian (Farsi) | Polish | Portuguese (Brazil) | Portuguese (Portugal) | Punjabi (Gurmukhi) | Romanian | Russian | Serbian (Cyrillic) | Slovak | Slovenian | Spanish | Swahili | Swedish | Tagalog (Filipino) | Tamil | Telugu | Thai | Turkish | Ukrainian | Urdu | Vietnamese
Prefer to Clone Locally?
This repository includes 50+ language translations which significantly increases the download size. To clone without translations, use sparse checkout:
Bash / macOS / Linux:
git clone --filter=blob:none --sparse https://github.com/microsoft/mcp-for-beginners.git cd mcp-for-beginners git sparse-checkout set --no-cone '/*' '!translations' '!translated_images'CMD (Windows):
git clone --filter=blob:none --sparse https://github.com/microsoft/mcp-for-beginners.git cd mcp-for-beginners git sparse-checkout set --no-cone "/*" "!translations" "!translated_images"This gives you everything you need to complete the course with a much faster download.
Welcome to your journey into the Model Context Protocol! If you've ever wondered how AI applications communicate with different tools and services, you're about to discover the elegant solution that's transforming how developers build intelligent systems.
Think of MCP as a universal translator for AI applications - just like how USB ports let you connect any device to your computer, MCP lets AI models connect to any tool or service in a standardized way. Whether you're building your first chatbot or working on complex AI workflows, understanding MCP will give you the power to create more capable and flexible applications.
This curriculum is designed with patience and care for your learning journey. We'll start with simple concepts you already understand and gradually build your expertise through hands-on practice in your favorite programming language. Every step includes clear explanations, practical examples, and plenty of encouragement along the way.
By the time you complete this journey, you'll have the confidence to build your own MCP servers, integrate them with popular AI platforms, and understand how this technology is reshaping the future of AI development. Let's begin this exciting adventure together!
The current protocol revision is MCP Specification 2026-07-28. The MCP specification uses date-based versioning (YYYY-MM-DD format) to make protocol compatibility explicit.
Version note: this curriculum teaches the current
2026-07-28concepts, including stateless requests, the Extensions framework, and the deprecation of Roots, Sampling, and Logging. Some hands-on examples remain explicitly versioned to2025-11-25while SDK support catches up. See What's Changed in MCP: The 2026-07-28 Specification for the changes and migration guidance.
These resources become more valuable as your understanding grows, but don't feel pressured to read everything immediately. Start with the areas that interest you most!
By the end of this curriculum, you'll feel confident and excited about your new abilities. Here's what you'll achieve:
• Understand MCP fundamentals: You'll grasp what the Model Context Protocol is and why it's revolutionizing how AI applications work together, using analogies and examples that make sense.
• Build your first MCP server: You'll create a working MCP server in your preferred programming language, starting with simple examples and growing your skills step by step.
• Connect AI models to real tools: You'll learn how to bridge the gap between AI models and actual services, giving your applications powerful new capabilities.
• Implement security best practices: You'll understand how to keep your MCP implementations safe and secure, protecting both your applications and your users.
• Deploy with confidence: You'll know how to take your MCP projects from development to production, with practical deployment strategies that work in the real world.
• Join the MCP community: You'll become part of a growing community of developers who are shaping the future of AI application development.
Before we dive into MCP specifics, let's make sure you feel comfortable with some foundational concepts. Don't worry if you're not an expert in these areas - we'll explain everything you need to know as we go!
Think of a protocol like the rules for a conversation. When you call a friend, you both know to say "hello" when you answer, take turns speaking, and say "goodbye" when you're done. Computer programs need similar rules to communicate effectively.
MCP is a protocol - a set of agreed-upon rules that help AI models and applications have productive "conversations" with tools and services. Just like how having conversation rules makes human communication smoother, having MCP makes AI application communication much more reliable and powerful.
You already use client-server relationships every day! When you use a web browser (the client) to visit a website, you're connecting to a web server that sends you the page content. The browser knows how to ask for information, and the server knows how to respond.
In MCP, we have a similar relationship: AI models act as clients that request information or actions, while MCP servers provide those capabilities. It's like having a helpful assistant (the server) that the AI can ask to perform specific tasks.
Imagine if every car manufacturer used different shaped gas pumps - you'd need a different adapter for each car! Standardization means agreeing on common approaches so things work together seamlessly.
MCP provides this standardization for AI applications. Instead of every AI model needing custom code to work with every tool, MCP creates a universal way for them to communicate. This means developers can build tools once and have them work with many different AI systems.
Your MCP journey is carefully structured to build your confidence and skills progressively. Each phase introduces new concepts while reinforcing what you've already learned.
This is where your adventure begins! We'll introduce you to MCP concepts using familiar analogies and simple examples. You'll understand what MCP is, why it exists, and how it fits into the larger world of AI development.
• Module 0 - Introduction to MCP: We'll start by exploring what MCP is and why it's so important for modern AI applications. You'll see real-world examples of MCP in action and understand how it solves common problems developers face.
• Module 1 - Core Concepts Explained: Here you'll learn the essential building blocks of MCP. We'll use plenty of analogies and visual examples to make sure these concepts feel natural and understandable.
• Module 2 - Security in MCP: Security might sound intimidating, but we'll show you how MCP includes built-in safety features and teach you best practices that protect your applications from the start.
Now the real fun begins! You'll get hands-on experience building actual MCP servers and clients. Don't worry - we'll start simple and guide you through every step.
This module includes multiple hands-on guides that let you practice in your preferred programming language. You'll create your first server, build a client to connect to it, and even integrate with popular development tools like VS Code.
Each guide includes complete code examples, troubleshooting tips, and explanations of why we make specific design choices. By the end of this phase, you'll have working MCP implementations you can be proud of!
With the basics mastered, you're ready to explore more sophisticated MCP features. We'll cover practical implementation strategies, debugging techniques, and advanced topics like multi-modal AI integration.
You'll also learn how to scale your MCP implementations for production use and integrate with cloud platforms like Azure. These modules prepare you to build MCP solutions that can handle real-world demands.
The final phase focuses on joining the MCP community and specializing in areas that interest you most. You'll learn how to contribute to open-source MCP projects, implement advanced authentication patterns, and build comprehensive database-integrated solutions.
Module 11 deserves special mention - it's a complete 13-lab hands-on learning path that teaches you to build production-ready MCP servers with PostgreSQL integration. It's like a capstone project that brings together everything you've learned!
| Module | Topic | Description | Link |
|---|---|---|---|
| Module 0-3: Fundamentals | |||
| 00 | Introduction to MCP | Overview of the Model Context Protocol and its significance in AI pipelines | Read more |
| 01 | Core Concepts Explained | In-depth exploration of core MCP concepts | Read more |
| 1.1 | What's Changed in MCP (2026-07-28) | Stateless protocol, Extensions framework, and feature deprecations in the current specification | Guide |
| 02 | Security in MCP | Security threats and best practices | Read more |
| 2.1 | CIMD and DCR Authorization | Compare preferred CIMD registration with deprecated DCR fallback using a protected MCP server | Sample |
| 03 | Getting Started with MCP | Environment setup, basic servers/clients, integration | Read more |
| Module 3: Building Your First Server & Client | |||
| 3.1 | First Server | Create your first MCP server | Guide |
| 3.2 | First Client | Develop a basic MCP client | Guide |
| 3.3 | Client with LLM | Integrate large language models | Guide |
| 3.4 | VS Code Integration | Consume MCP servers in VS Code | Guide |
| 3.5 | stdio Server | Create servers using stdio transport | Guide |
| 3.6 | HTTP Streaming | Implement HTTP streaming in MCP | Guide |
| 3.7 | Microsoft Foundry Toolkit | Use Microsoft Foundry Toolkit with MCP | Guide |
| 3.8 | Testing | Test your MCP server implementation | Guide |
| 3.9 | Deployment | Deploy MCP servers to production | Guide |
| 3.10 | Advanced server usage | Use advanced servers for advanced feature usage and improved architecture | Guide |
| 3.11 | Simple auth | A chapter showing you auth from the beginning and RBAC | Guide |
| 3.12 | MCP Hosts | Configure Claude Desktop, Cursor, Cline, and other MCP hosts | Guide |
| 3.13 | MCP Inspector | Debug and test MCP servers with the Inspector tool | Guide |
| 3.14 | Sampling | Use sampling to collaborate with the client | Guide |
| 3.15 | MCP Apps | Build MCP Apps | Guide |
| Module 4-5: Practical & Advanced | |||
| 04 | Practical Implementation | SDKs, debugging, testing, reusable prompt templates | Read more |
| 4.1 | Pagination | Handle large result sets with cursor-based pagination | Guide |
| 05 | Advanced Topics in MCP | Multi-modal AI, scaling, enterprise use | Read more |
| 5.1 | Azure Integration | MCP Integration with Azure | Guide |
| 5.2 | Multi-modality | Working with multiple modalities | Guide |
| 5.3 | OAuth2 Demo | Implement OAuth2 authentication | Guide |
| 5.4 | Root Contexts | Understand and implement root contexts | Guide |
| 5.5 | Routing | MCP routing strategies | Guide |
| 5.6 | Sampling | Sampling techniques in MCP | Guide |
| 5.7 | Scaling | Scale MCP implementations | Guide |
| 5.8 | Security | Advanced security considerations | Guide |
| 5.9 | Web Search | Implement web search capabilities | Guide |
| 5.10 | Realtime Streaming | Build realtime streaming functionality | Guide |
| 5.11 | Realtime Search | Implement realtime search | Guide |
| 5.12 | Entra ID Auth | Authentication with Microsoft Entra ID | Guide |
| 5.13 | Foundry Integration | Integrate with Microsoft Foundry | Guide |
| 5.14 | Context Engineering | Techniques for effective context engineering | Guide |
| 5.15 | MCP Custom Transport | Custom Transport implementations | Guide |
| 5.16 | Protocol Features | Progress notifications, cancellation, resource templates | Guide |
| 5.17 | Adversarial Multi-Agent Reasoning | Two agents argue opposite sides using shared MCP tools, evaluated by a judge agent | Guide |
| Module 6-10: Community & Best Practices | |||
| 06 | Community Contributions | How to contribute to the MCP ecosystem | Guide |
| 07 | Insights from Early Adoption | Real-world implementation stories | Guide |
| 08 | Best Practices for MCP | Performance, fault-tolerance, resilience | Guide |
| 09 | MCP Case Studies | Practical implementation examples | Guide |
| 10 | Hands-on Workshop | Building an MCP Server with Microsoft Foundry Toolkit | Lab |
| Module 11: MCP Server Hands On Lab | |||
| 11 | MCP Server Database Integration | Comprehensive 13-lab hands-on learning path for PostgreSQL integration | Labs |
| 11.1 | Introduction | Overview of MCP with database integration and retail analytics use case | Lab 00 |
| 11.2 | Core Architecture | Understanding MCP server architecture, database layers, and security patterns | Lab 01 |
| 11.3 | Security & Multi-Tenancy | Row Level Security, authentication, and multi-tenant data access | Lab 02 |
| 11.4 | Environment Setup | Setting up development environment, Docker, Azure resources | Lab 03 |
| 11.5 | Database Design | PostgreSQL setup, retail schema design, and sample data | Lab 04 |
| 11.6 | MCP Server Implementation | Building the FastMCP server with database integration | Lab 05 |
| 11.7 | Tool Development | Creating database query tools and schema introspection | Lab 06 |
| 11.8 | Semantic Search | Implementing vector embeddings with Azure OpenAI and pgvector | Lab 07 |
| 11.9 | Testing & Debugging | Testing strategies, debugging tools, and validation approaches | Lab 08 |
| 11.10 | VS Code Integration | Configuring VS Code MCP integration and AI Chat usage | Lab 09 |
| 11.11 | Deployment Strategies | Docker deployment, Azure Container Apps, and scaling considerations | Lab 10 |
| 11.12 | Monitoring | Application Insights, logging, performance monitoring | Lab 11 |
| 11.13 | Best Practices | Performance optimization, security hardening, and production tips | Lab 12 |
| Module 12: MCP Tooling | |||
| 12.1 | Tooling | MCP usage in Copilot App | Guide |
One of the most exciting parts of learning MCP is seeing your code skills develop progressively. We've designed our code examples to start simple and grow more sophisticated as your understanding deepens. Here's how we introduce concepts - with code that's easy to understand but demonstrates real MCP principles, you'll understand not just what this code does, but why it's structured this way and how it fits into larger MCP applications.
| Language | Description | Link |
|---|---|---|
| C# | MCP Server Example | View Code |
| Java | MCP Calculator | View Code |
| JavaScript | MCP Demo | View Code |
| Python | MCP Server | View Code |
| TypeScript | MCP Example | View Code |
| Rust | MCP Example | View Code |
| Language | Description | Link |
|---|---|---|
| C# | Advanced Sample | View Code |
| Java with Spring | Container App Example | View Code |
| JavaScript | Advanced Sample | View Code |
| Python | Complex Implementation | View Code |
| TypeScript | Container Sample | View Code |
To get the most out of this curriculum, you should have:
Basic knowledge of programming in at least one of the following languages: C#, Java, JavaScript, Python, or TypeScript
Understanding of client-server model and APIs
Familiarity with REST and HTTP concepts
(Optional) Background in AI/ML concepts
Joining our community discussions for support
This repository includes several resources to help you navigate and learn effectively:
A comprehensive Study Guide is available to help you navigate this repository effectively. This visual curriculum map shows how all the topics connect and provides guidance on how to use the sample projects effectively. It's especially helpful if you're a visual learner who likes to see the big picture.
The guide includes:
We maintain a detailed Changelog that tracks all significant updates to the curriculum materials, so you can stay current with the latest improvements and additions.
Each lesson in this guide includes:
Congratulations! You've just taken the first step in an exciting journey that will expand your programming capabilities and connect you to the cutting edge of AI development.
By reading through this introduction, you've already begun building your MCP knowledge foundation. You understand what MCP is, why it matters, and how this curriculum will support your learning journey. That's a significant achievement and the beginning of your expertise in this important technology.
As you progress through the modules, remember that every expert was once a beginner. The concepts that might seem complex now will become second nature as you practice and apply them. Each small step builds toward powerful capabilities that will serve you throughout your development career.
You're joining a community of learners and experts who are passionate about MCP and eager to help others succeed. Whether you're stuck on a coding challenge or excited to share a breakthrough, the community is here to support your journey.
If you get stuck or have any questions about building AI apps. Join fellow learners and experienced developers in discussions about MCP. It's a supportive community where questions are welcome and knowledge is shared freely.
If you have product feedback or errors while building visit:
Your MCP adventure starts now! Begin with Module 0 to dive into your first hands-on MCP experiences, or explore the sample projects to see what you'll be building. Remember - every expert started exactly where you are now, and with patience and practice, you'll be amazed at what you can achieve.
Welcome to the world of Model Context Protocol development. Let's build something amazing together!
This curriculum grows stronger with contributions from learners like you! Whether you're fixing a typo, suggesting a clearer explanation, or adding a new example, your contributions help other beginners succeed.
Thanks to Microsoft Valued Professional Shivam Goyal for contributing code samples.
The contribution process is designed to be welcoming and supportive. Most contributions require a Contributor License Agreement (CLA), but the automated tools will guide you through the process smoothly.
This entire curriculum is available under the MIT LICENSE, meaning you can use, modify, and share it freely. This supports our mission of making MCP knowledge accessible to developers everywhere.
This project welcomes contributions and suggestions. Most contributions require you to agree to a Contributor License Agreement (CLA) declaring that you have the right to, and actually do, grant us the rights to use your contribution. For details, visit .
When you submit a pull request, a CLA bot will automatically determine whether you need to provide a CLA and decorate the PR appropriately (e.g., status check, comment). Simply follow the instructions provided by the bot. You will only need to do this once across all repos using our CLA.
This project has adopted the Microsoft Open Source Code of Conduct. For more information see the Code of Conduct FAQ or contact [email protected] with any additional questions or comments.
Ready to start your MCP journey? Begin with Module 00 - Introduction to MCP and take your first steps into the world of Model Context Protocol development!
Our team produces other courses! Check out:
内容由编辑整理，部分信息来自 GitHub 与项目官网，版权归原作者所有。发现错误？报告问题
