A 424-page textbook for building agentic systems

What it does — This repository hosts the complete materials for Antonio Gulli’s Agentic Design Patterns, a 424-page textbook on building intelligent AI systems. It includes the full PDF plus a Jupyter notebook for each of the 21 chapters and 7 appendices, covering everything from basic prompt chaining to inter-agent communication protocols. This is a curriculum, not a framework: you read the chapter, then run the matching notebook.

The interesting bit — Most agent repositories ship black-box frameworks; this one ships a taxonomy. By treating patterns like routing, reflection, and resource-aware optimization as first-class architectural concepts, it forces you to think about why agents are structured a certain way before copying code. The inclusion of newer protocol-level topics like Model Context Protocol (MCP) and Agent-to-Agent (A2A) alongside classic RAG and guardrail patterns is unusual for educational material.

Key highlights —

• 21 chapters plus 7 appendices spanning core, advanced, production, and enterprise patterns.
• Each chapter has a dedicated Jupyter notebook; the PDF is included in full under book/.
• Covers practical production concerns: exception handling, human-in-the-loop, evaluation, safety guardrails, and resource-aware optimization.
• Code examples are MIT-licensed; the book text itself is copyrighted educational material.
• All author royalties are donated to Save the Children.

Caveats —

• This is educational companion content, not an importable library or actively maintained framework.
• The README lists generic dependencies but does not clarify which specific frameworks the notebooks rely on.

Verdict — A solid starting point if you want to learn agent architecture from first principles with theory and code side-by-side. Look elsewhere if you need a drop-in framework or a purely code-driven quickstart.