The Agent Course Roadmap Shows Which Lessons to Read

Want to understand AI agents systematically? Microsoft's open-source AI Agents for Beginners is a solid foundation. We rewrote its 19 lessons in plain language, and each lesson has key points, hands-on exercises, and a tutor prompt you can hand directly to an agent: you can read it yourself, and an agent can guide you through it.

First, figure out which reader you are. The course was originally written for developers: it teaches you to build agents with code. So we split the 19 lessons into two groups:

• Readable for general users (5 lessons): Covers concepts. The exercises can be done in a chat window without writing code.
• Advanced for developers (14 lessons): Requires writing code, designing data structures, and deployment. Suited to people who want to build their own agents. If you don't write code, it's completely fine to skip this group.

For General Users

Lesson	Topic	In one line
1	What an agent really is, and which tasks are worth using one for	What an agent is, and which tasks are worth it
5	Agentic RAG: if you can't find it, change your approach and search again	If nothing is found, rephrase the question and search again
9	Metacognition: getting the agent to check its own work	Get the agent to check its own work
12	Context engineering: don't overfill the window	Manage what the model can see right now
13	Memory: getting the agent to remember you	Get the agent to remember you

For Advanced Developers

Lesson	Topic	In one line
0	Set up your environment first so the course code can run	What you need to run the course code
2	Choosing a framework: don't reinvent the wheel	What frameworks save you from
3	Three principles for designing agents: transparent, controllable, consistent	The three principles: transparent, controllable, consistent
4	Tool use: letting the model actually take action	Function calling lets the model act
6	Is an agent trustworthy? Start with these threat types	Six threat types and their defenses
7	Planning: break the task down before starting	Break tasks down before starting
8	Multi-agent: when to bring in a few helpers	When to bring in more helpers
10	After launch: how to see what it's really doing	After launch, how to see what it's doing
11	MCP, A2A, NLWeb: what each of the three protocols covers	The division of labor among MCP, A2A, and NLWeb
14	What Microsoft Agent Framework can do for you	What Microsoft Agent Framework provides
15	Browser agents: getting things done online like a person	Getting things done online like a person
16	From laptop to production: what else an agent needs	What to add from laptop to production
17	Running agents locally: your data stays home	Running a small-model agent on your own machine
18	Leaving a tamper-proof ledger for agent actions	A tamper-proof ledger for actions

Suggested Order

• General users: 01 → 05 → 12 → 13 → 09. Finish these and you'll be better at using ready-made agents.
• Developers: Read the 5 lessons above first, then work through the advanced group in order from 00 to 18. Lesson 00 covers running the original code and requires an Azure account.

References

• Original project: microsoft/ai-agents-for-beginners, MIT license
• This series is a plain-language rewrite; for details and the latest content, refer to the original project.