Mastering AI Agents: The 10 Best Free Courses, Tutorials & Learning Tools
2025 is the year of the agents. Make it the year you boss them around.
Almost all the tech giants are betting on AI agents: Nvidia, OpenAI, Microsoft, Google, and even companies that haven’t developed much AI before, as Salesforce or SAP. Mark Zuckerberg thinks eventually there will be more . So, speaking of our planet, how on earth do you learn the basic techniques of agents?
There are many expensive courses on the web and even more expensive certificates.
My recommendation: Don’t pay.
I’ve rounded up some great courses and resources for you that are free.
You will need about a weekend to work through all of it. With a little bit of experimentation and coding, maybe a bit longer.
To provide a little guidance, I decorate the contents with martial arts-inspired belt colors based on complexity and target audience.
⬜⬜⬜⬜⬜ Intuitive
🟩🟩🟩🟩🟩 Intermediate
🟥🟥🟥🟥🟥 Advanced
🟫🟫🟫🟫🟫 Expert
⬛⬛⬛⬛⬛ Master
And now, go, fight!
#1 AI Agents Explained: A Short Video Guide
⬜⬜⬜⬜⬜ intuitive
How are AI agents structured, what can they do? How do they compare to traditional models and software? How can they use dynamic data? AI Alfie explains it in just 7 minutes. Alfie talks faster than I can read.
#2: What Are AI Agents: A Step-by-Step Guide to Build Your Own
⬜⬜⬜⬜⬜ 🟥🟥🟥🟥🟥 intuitive -> advanced
I really spent a lot of time in this story explaining what an AI agent is and how to design a commercial agent on paper, with all the complexities that arise in the real world:
- how to bring together algorithms and AI
- how do I classify user input
- how do I extract data
- how do I make the data persistent
- how do I get the accuracy to a reasonable level
- and when do I have to handover to the human team
Here is a by Anthropic, detailing some of the topics a little bit further.
#3: Six AI Agents So Good, They Feel Illegal
⬜⬜⬜⬜⬜⬜ intuitive
Just to warm up: Mohit Vaswani lists 6 agents that really melt the butter on the bread. Check them out, they are examples of what you can do with agentic AI. Oh, and at least one of the companies behind the agents is a unicorn — with a valuation north of $1B. Maybe that fact will make you want to get started yourself.
#4: Free Video Course on DeepLearning.ai: Building a Small Agent System with CrewAI
🟩🟩🟩🟩🟩 🟥🟥🟥🟥🟥 intermediate -> advanced
Multi AI Agent Systems sounds like you’re about to command an army of futuristic AI warriors — but relax, it’s just a bunch of workflow nodes doing their job. You don’t need a PhD in rocket science to learn & build. With CrewAI and João’s instructions, you can spin up your own little agent system in no time. And if needed for real-world production-grade software you can move on to fancier frameworks.
Here is the link where you can start the course directly (login required, but no payment):
And no, it is not just blah-blah, but a lot of code with a Jupyter notebook to try things out:
#5: Agentic AI Tool & Stack Overviews
🟩🟩🟩🟩🟩 🟥🟥🟥🟥🟥 intermediate -> advanced
How do I get started and move forward? What should I use as models, frameworks, hosting, tracking, tools, vector DBs?
Letta not only makes , but also describes some of the individual components of the stack. Oh, and they are also on it themselves — with pleasure!
Chiefmartec gives a good overview on agents, builders, frameworks.
For those of us who need something more vibrant and colorful (a.k.a. me, after a long day at work), check out 360DigiTMG’s post:
#6: AI Agents vs. Traditional Generative AI vs. RPA
⬜⬜⬜⬜⬜ 🟩🟩🟩🟩🟩 intuitive -> intermediate
How does agentic AI compare to traditional generative AI? To robotic process automation? Is a chatbot an agent? What does it need to be an agent?
that does an excellent job of detailing a very rough sketch I once used to try to map the world of agentic AI.
Cobus Greyling gives a great introduction to the specific architectures and components of agentic AI systems:
Here, Tyler McGregory compares :
#7. AI Agent Frameworks & Comparisons
🟩🟩🟩🟩🟩 🟫🟫🟫🟫🟫 intermediate -> expert
Major frameworks to build AI agents comprise:
- LlamaIndex ,
- HuggingFace’
- Microsoft’s
- Haystack’s
- Pydantic
- OpenAI’s
Yi Zhang with a lot of code snippets and graphic footage. Great!
Here is cool story from Aparna Dhinakaran, same question, different take: Choosing between Agent Frameworks (comparing no framework, LangGraph, LlamaIndex Workflows) with code examples.
Here is a very broad and multifaceted comparison of all the frameworks mentioned above plus by MA Raza, Ph.D.
#8. Free AI Agent Course with Certificate on Hugging Face
🟩🟩🟩🟩🟩 🟫🟫🟫🟫🟫 intermediate -> expert
is HuggingFace’s cute introduction to the topic with background concepts and code snippets.
It covers, among other things:
- Tool Usage
- Thought-Action-Observation Cycle
- Thought, Internal Reasoning and the Re-Act Approach
- A Dummy Agent Library
They use their own agent framework as well as LlamaIndex and LangChain.
The course includes small test sections and a final test with certificate:
The first unit is already available, and they are working on the others.
#9: Multi Agentic RAG with Hugging Face Code Agents
🟫🟫🟫🟫🟫 ⬛⬛⬛⬛⬛ expert -> advanced
Enough talk. Now, you want to dive deep into coding, you want to implement something. Here is a long, detailed story with comprehensive explanations on how to set up a multi-agent system by Gabriele Sgroi, PhD. With inline code, explanations, GIT repository, notebook, bed and maybe even breakfast.
#10: Sam Altman & Team Introducing AI Agents and Their Operator Agent
⬜⬜⬜⬜⬜ 🟩🟩🟩🟩🟩 intuitive -> intermediate
Sam and his team introduce the OpenAI Operator, an agent that can work with the browser, i.e. to open URLs and interact with the outside world.
I like the case at 5:30 min: Throw in a handwritten shopping list and Operator just clicks through webpages and buys the stuff online. It can really “see” and “read” a webpage, select buttons and click them, select fields and fill them based on its task.
#11: Bonus Track: How I Got a Big AI Agent Up and Running — What Worked and What Didn’t
🟥🟥🟥🟥🟥 ⬛⬛⬛⬛⬛ advanced -> master
I can throw this in as a bonus — because it’s from me - personally: The lessons I learned in my last (pretty big) agent projects, like:
- Fact Over Fiction: The Battle for Accuracy
- Be Ambitious: 80% and More Fully Automated End-to-end Processing of Complex Tasks is Clearly Possible
- Convince Stakeholder, Partners, Clients with a KPI Based Agreement
- Low Code, No-Code is Nice, But Not for Production
… and 9 more lessons.
So, folks, that’s it. Use the resources, learn, learn, learn until your synapses glow. Try out the examples. Start coding agents.
I myself feel that I have learned more in the last 5 years, since my personal start in generative AI, than in all the years before.
So, let’s use the momentum and ride the wave. 🏄
And if you feel like it:
Follow me on Medium (⇈) or for updates and new stories about generative AI, AI agents, and prompt engineering.
(Many thanks to Almudena and Kirsten! Title image created with Midjourney AI)
