microsoft/ai-agents-for-beginners

18 Lessons to Get Started Building AI Agents

Summary

microsoft/ai-agents-for-beginners is an 18-lesson course for learning how to build AI agents, with lessons covering agent frameworks, design patterns, tools, retrieval, memory, protocols, deployment, local agents, and security. Each lesson combines written material, Python code samples, and links to further learning.

What it is useful for

The course is useful for engineers who need a structured introduction to agent development or a shared vocabulary for discussing agent architecture. Teams can use individual lessons as onboarding material or as references while designing their own systems.

Concrete topics include:

• Tool use and planning for agents that take actions
• Agentic RAG for retrieval-oriented workflows
• Multi-agent, metacognition, context-engineering, and memory patterns
• MCP, A2A, and NLWeb protocols
• Trustworthy, production, deployment, local-agent, and security concerns

It also supports engineers who want more context before working with generative models. The README recommends the separate Generative AI for Beginners course for people building with generative AI for the first time.

The course is available in more than 50 languages, which can help distributed teams study in their preferred language. However, the translations and translated images significantly increase clone size.

How engineers can use it

Start with the lesson that matches the problem you need to solve rather than following the sequence strictly. For example, begin with Tool Use, Agentic RAG, or Planning Design Pattern for an application being designed; use AI Agents in Production, Deploying Scalable Agents, or Securing AI Agents when moving beyond an initial prototype.

The exercises are in the code_samples directory. Fork the repository to create your own copy, then follow 00-course-setup/README.md for environment and provider configuration. The supplied README does not document the complete dependency-installation and configuration procedure, so it is the appropriate starting point rather than inferring setup commands.

If you only need the English course and want a smaller download, the README documents this sparse checkout:

git clone --filter=blob:none --sparse https://github.com/microsoft/ai-agents-for-beginners.git
cd ai-agents-for-beginners
git sparse-checkout set --no-cone '/*' '!translations' '!translated_images'

Most exercises use Microsoft Agent Framework with Microsoft Foundry Agent Service V2, and Microsoft Foundry requires an Azure account. Some samples also support alternative OpenAI-compatible providers such as MiniMax, but that support is limited to certain code samples.