GenAI Agents

What is inside

The notebooks are organised by category:

• Beginner-friendly agents: a simple conversational agent, question answering, basic data analysis.
• Framework tutorials: the same ideas in LangGraph, and examples with other frameworks such as CrewAI, AutoGen, OpenAI Swarm and PydanticAI.
• Educational and research agents: academic assistants and a scientific paper agent.
• Business and professional agents: customer support, and other workplace tasks.
• Creative and content agents, analysis agents, news agents, shopping agents and productivity agents.
• Advanced techniques: controllable RAG agents, human-in-the-loop approval, and trace-based evaluation.

LangGraph is the main framework across most notebooks.

The repository had about 24,000 GitHub stars as of September 2026 and was updated that month. It uses a custom non-commercial licence.

Why a collection beats a single tutorial

A single tutorial teaches you one author's design choices as if they were the only ones. Reading several implementations side by side shows you the real design space: when an agent needs a planner, when a simple router is enough, how state is structured, where a human approval step goes. That comparison is how you develop judgement about agent design.

How to use it well

1. Start with two beginner notebooks and run them end to end. Note how the graph, state and tools are defined.
2. Pick one intermediate agent close to a problem you care about and rebuild it from scratch without copying, using the notebook only when stuck.
3. Compare two agents in the same category and write down the design differences and why they might matter.
4. Read the advanced notebooks on human-in-the-loop and evaluation before building anything others will use.

Keep an eye on library versions. LangGraph and other frameworks change quickly, so some notebooks may need small updates.

What it does not cover

It is about agent design, not production. Deployment, memory at scale, security, monitoring and cost control are covered by the same author's Agents Towards Production repository. Evaluation methodology is better learned from a dedicated resource such as Hamel Husain and Shreya Shankar's evals FAQ.

Before you start

You need solid Python, an API key for a model provider, and ideally a first exposure to agent concepts from a structured course such as Microsoft's AI Agents for Beginners.

Common questions

Is it suitable for beginners? The beginner notebooks are, if you already know Python. Many others assume familiarity with LangGraph.

Can I use these agents in a product? Check the licence first; it is non-commercial.

Best for: developers who have built one agent and want to see many different designs to learn how experienced builders structure them.