The Hugging Face Agents Course: Free AI Agents Curriculum
/agents-course
agents-course provides a structured, framework-neutral path into AI agent development, with its strength being the final benchmark project where actual agent performance is evaluated. The course offers comfort with smolagents, LangGraph, and LlamaIndex, but not mastery. It is worth starting for those who learn best with a guided curriculum, and worth skipping when framework-specific specialization is needed.
What is the Hugging Face Agents Course?
The Hugging Face Agents Course is a free, four-unit curriculum that takes you from LLM fundamentals to building, evaluating, and benchmarking AI agents. It spans smolagents, LangGraph, and LlamaIndex, includes Agentic RAG, and ends with a certified final project scored against a public leaderboard. Content lives on the Hugging Face learning platform with no local setup required.
Who Should Take This Course?
Python developers with basic LLM exposure who want a guided path into agentic AI before diving into a specific framework. Backend engineers, data engineers, and full-stack developers will get the most value β absolute beginners should build Python fundamentals first.
Key Learning Modules
- βUnit 1 introduces core agent concepts: what agents are, how LLMs drive them, model family trees, and special tokens that control tool use and reasoning
- βUnit 2 walks through three frameworks β smolagents for lightweight agent creation, LangGraph for controlled production flows, and LlamaIndex for data-indexed agents
- βUnit 2 Bonus covers observability and evaluation, teaching you to trace agent decisions and measure performance reliably
- βUnit 3 focuses on Agentic RAG, using agents to route queries and pull from dynamic data sources rather than static retrievers
- βUnit 4 is a final project with automated evaluation and a public leaderboard β your agent is benchmarked, not just graded
- βBonus units include fine-tuning an LLM for function-calling and a Pokemon-based game agent project for hands-on variety
Why Learn About AI Agents Now?
Agentic AI has moved from research demos to production systems, with models now capable of multi-step reasoning, tool use, and autonomous task completion at a level that wasn't viable before. That shift creates genuine demand for developers who understand how to design agent loops, manage toolchains, and evaluate reliability β skills the job market is actively asking for now. The Hugging Face Agents Course is part of that moment, giving learners a credentialed way to signal they understand the space.
Strengths
- βThe course offers a comprehensive curriculum, progressing from the basics of AI agents and LLMs to advanced topics like agentic RAG and a final project with automated evaluation.
- βIt provides detailed introductions to several prominent AI agent frameworks, including `smolagents`, `LlamaIndex` for data interaction, and `LangGraph` for controlling agent workflows.
- βIncludes practical bonus units covering essential skills such as fine-tuning LLMs for function-calling, implementing observability and evaluation for agents, and exploring agent use cases in games like PokΓ©mon.
- βThe entire course content is freely accessible online through the Hugging Face platform, making it an open resource for learning about AI agents.
- βThe project encourages community involvement, welcoming contributions for typo fixes and new units, with guidelines for proposing content.
Course Scope and Prerequisites
- β³Requires existing foundational knowledge in Python and Large Language Models (LLMs), so it's not suitable for absolute beginners in those areas.
- β³The course focuses on specific agent frameworks like `smolagents`, `LlamaIndex`, and `LangGraph`, and doesn't aim to cover every available AI agent framework.
- β³While comprehensive for its scope, the curriculum is structured from basics to a final project, and may not delve into highly advanced or experimental agent research topics.
- β³Adding entirely new units or significant content requires creating an issue and discussion with the course maintainers, indicating a curated rather than fully open contribution model for major changes.
Other Ways to Learn About AI Agents
Frequently Asked Questions
Yes, the course is completely free. You can access all units, bonus materials, and the final benchmark project at no cost through the Hugging Face learning platform.
You need basic Python knowledge and a foundational understanding of how LLMs work. If you've used an LLM API or built a prompt before, you have enough background to follow along.
The curriculum covers agent fundamentals, smolagents, LangGraph, LlamaIndex, Agentic RAG, observability, agent evaluation, a final benchmark project, and bonus units on fine-tuning function-calling models and game agents.
For small fixes like typos, submit a pull request directly. For new units, open an issue describing the topic and rationale first β the maintainers review proposals before inviting collaboration.
The course covers three frameworks in depth: smolagents for lightweight agents, LangGraph for production-controlled flows, and LlamaIndex for data-indexed agents. Observability and evaluation round out the tooling.
Use the provided BibTeX entry in the README, listing authors Burtenshaw, Thomas, Simonini, and Paniego, the 2025 publication year, and the GitHub repository URL as the source.
Best use cases
- β’Learning the core concepts of AI agents, including definitions, LLMs, and special tokens, for developers new to the field.
- β’Building AI agents using popular frameworks like `smolagents`, `LlamaIndex` for data interaction, and `LangGraph` for production-ready applications.
- β’Fine-tuning Large Language Models specifically for function-calling capabilities, a practical skill for custom agent development.
- β’Implementing Agentic RAG (Retrieval Augmented Generation) to enhance agents' ability to respond to varied use cases by leveraging external data.
- β’Evaluating and testing AI agents through automated methods, including a final project focused on certification and performance benchmarking.
Related repositories
Still deciding about agents-course?
One click hands the question to an AI along with this page β see what it says about agents-course.