# Entry
They are all construction agents. Far fewer people can accurately explain why their agent loops endlessly, ignores a tool they’ve been given, or confidently reports success on a task they never completed. The difference between sending an agent and understanding it is that there are five resources there, and each of them is completely free.
I’ve deliberately mixed up the registers here: a practical course that can be completed in a weekend, a strict academic text once the noise dies down and the basics are needed, and a few things in between. Analyze even three of them and you’ll design agents like someone who knows what’s going on in the orchestration, not someone who pastes in hints and hopes.
# AI Agents for Beginners (Microsoft)
Start here if you want structure. AI agents for beginners is a full course on GitHub under the MIT license, with over fifteen lessons with video instructions and runable Python for each lesson. It goes from the real fundamentals – what an agent is and when you actually need one – through design patterns you’ll constantly reuse: tool usage, scheduling, search-aided generation (RAG), multi-agent configurations, and memory and context engineering that separate a demo from something useful.
What makes it the best free starting point is that it is maintained, not abandoned, and includes newer interoperability standards such as Model Context Protocol (MCP), which most 2023-era material is entirely older. This is closest to a structured manual that also compiles itself.
# Face Hugging AI Agents Course
The Face Hugging Agents Course is the one to pair with Microsoft because it is relentlessly practical and compares frameworks. You’re building agents spanning smolagents, LlamaIndex, and LangGraph, rather than combining a single library, and that’s exactly the perspective you want before committing your production stack to a single ecosystem.
It’s truly free, there’s no paid tier, and it ends with a benchmark project and certificate, so there’s a finish line rather than a never-ending playlist. If a Microsoft course teaches you concepts, this one will give you calluses.
# Building effective agents (anthropic)
Anthropic Engineering Guide Building effective agents it’s compact and that’s the point. He draws the most useful distinction in the field – between workflows (immense language models that follow predefined paths) and agents (immense language models that drive their own process) – and then catalogs a handful of patterns worth knowing: speedy chaining, routing, parallelism, coordinator-worker, and evaluator-optimizer loops.
His best contribution is a warning that most tutorials ignore: agents involve higher costs and the potential for error compounding, so you should go for the simplest thing that works and only add autonomy when the problem requires it. Read it when your first agent misbehaves and you feel like someone is explaining your own mistake to you.
# Multi-agent systems (Shoham and Leyton-Brown)
When the noise dies down and you want to know why multi-agent systems behave the way they do, Multi-agent systems by Yoav Shoham and Kevin Leyton-Brown is a strict foundation. The authors, with the publisher’s consent, provide a free electronic copy; download it from this site instead of looking for the PDF file elsewhere, as they explicitly ask readers to link to the source.
This is game theory, distributed decision-making, and the logical basis for today’s agent conversations. It predates the era of immense language models, which is why it is so valuable: the problems of coordination, negotiation, and motivation between agents are aged and well-researched, and most people who would rediscover them now would save weeks if they read the theory itself once.
# A series of whitepapers from Google and Kaggle agents
# Where to go next
Nahla Davies is a programmer and technical writer. Before devoting herself full-time to technical writing, she managed, among other intriguing things, to serve as lead programmer for a 5,000-person experiential branding organization whose clients include: Samsung, Time Warner, Netflix and Sony.