Designing AI agents is the art of creating systems that can perceive information, reason, make decisions, and act autonomously. This book unveils a practical approach to building robust, manageable, and scalable agent systems based on modern large language models.
What This Book Is About
The material offers a systematic view of designing AI agents, integrating seven key cognitive functions — perception, memory, reasoning, action, reflection, collaboration, and governance. Each function is examined in conjunction with six architectural topologies, allowing for the creation of flexible and resilient agent solutions.
The author describes 27 proven design patterns that help create industrial agent systems capable of working effectively in complex conditions and solving applied tasks of various levels.
Practical Focus
Instead of theoretical discussions, the book is built around a single practical example — creating an agent for code review. Step by step, the reader observes how a simple prototype evolves into a fully-fledged production system.
Additional Cases
- automation of DevOps processes;
- compliance checking;
- research analytics for complex projects.
Who This Edition Is For
The book is aimed at developers who are already familiar with the basics of building agent systems and are looking to deepen their knowledge in AI agent engineering. Readers will gain a unified conceptual framework, learn about embedded governance principles, and get acquainted with standard development tools, including Claude Code, Cursor, and OpenClaw.
What You Will Learn
- how to choose the right architecture for an agent system;
- how to organize an agent's cognitive functions;
- how to design reliable and manageable AI agents;
- which approaches and patterns have proven effective in practice;
- which tools help speed up development and deployment.
