Building Agentic AI Systems: Build connected agent systems with MCP, A2A, agent loops, harness engineering, and cloud deployment, 2nd Edition
Anjanava Biswas, Wrick Talukdar
As AI agents move from experimental projects to real-world applications, developers are discovering that building an intelligent agent is only the beginning. The real challenge lies in making multiple agents work together, connect with different tools and platforms, handle failures, and deliver consistent results. Addressing these challenges requires more than powerful models or sophisticated frameworks.
This book explores how agents communicate, collaborate, and operate reliably, with a focus on building connected systems that are ready for production, exploring the practical aspects of engineering modern agentic AI systems. It demonstrates in detail how memory, context engineering, reasoning loops, and harness engineering make agents more capable and resilient. It then explores how protocols such as MCP and A2A enable agents to connect with tools and collaborate across systems. Designing multi-agent workflows and deploy distributed systems through frameworks such as LangGraph, CrewAI are explaind in detail.
The book also examines what it takes to keep agentic systems running effectively in production, exploring techniques for evaluating agent behavior, tracing and debugging workflows, managing costs, and addressing security, safety, and governance concerns. How Small Language Models (SLMs) can support more efficient deployments is discovered as well.














There are no reviews yet.