The Agentic Paradigm: Designing and building multi-agent applications that are reliable, secure, and responsible (English Edition)

Agentic AI has moved from research demos to the core of how modern software is built. Powered by language models, autonomous agents now reason, remember, use tools, and collaborate to solve problems no single model could-reshaping how applications are designed across every industry.

The Agentic Paradigm is a practitioner's guide to building multi-agent systems end to end. It begins with the anatomy of an agent-perception, memory, reasoning, and action, and shows how to build one using LangChain and LangGraph. It then covers multi-agent communication, coordination patterns, and emerging interoperability protocols, before applying these ideas to real use cases across finance, manufacturing, and healthcare. The closing chapters address what separates a demo from production: testing, observability, security, and building responsible systems aligned with ethical, social, and legal standards.

By the end, you will be able to design, architect, optimize, secure, and observe multi-agent systems and judge confidently when an agentic design is the right choice and when a classical one will serve better.

WHAT YOU WILL LEARN

● Build agents with LangChain and LangGraph.

● Understand perception, memory, reasoning, and action into agents.

● Design coordination in multi-agent systems with proven patterns and protocols.

● Apply agents to finance, manufacturing, and healthcare problems.

● Test, observe, secure, and govern agents for production.

WHO THIS BOOK IS FOR

This book is for software engineers, architects, and technical leads who want to build enterprise-grade agentic applications, as well as data scientists, ML practitioners, and researchers moving into agent development. A basic familiarity with Python and software design can be helpful for students, but no prior experience with agents is assumed.