Agentic AI Engineering: Build, secure, and operate production-grade AI agents with durable patterns, MCP, and A2A

Learn how to design, build, and operate production-grade AI agents with working OpenAI Agents SDK code, MCP and A2A integration, and enterprise controls for evaluation, security, observability, and cost

Key Features

• Learn architecture patterns that outlast model releases, proven in Fortune 500 deployments
• Orchestrate single-agent and multi-agent systems with the OpenAI Agents SDK, MCP, and A2A
• Catch failures before your users do with evaluation gates, guardrails, and tracing
• Purchase of the print or Kindle book includes a free PDF eBook

Book DescriptionSomewhere right now, an agent demo is going beautifully. Six months later, the same agent is looping on a malformed invoice, exposing data it should never have seen, and burning through its token budget. This book is about the distance between those two moments. You'll build single-agent and multi-agent systems with the OpenAI Agents SDK, applying production-ready patterns including planner/executor, router, supervisor/worker, tiered memory, tool registries, and human-in-the-loop workflows. You'll integrate enterprise systems through function calling and the Model Context Protocol (MCP), coordinate specialists with A2A, and tackle side effects, shared state, and failed tool calls. Then the focus shifts to what rarely appears in demos: evaluation harnesses, regression gates, prompt injection defenses, memory poisoning, tracing, cost visibility, versioned prompts, gradual rollouts, and tenant isolation. Written by Mohsen Amiribesheli, PhD, whose AI systems have served pharma, banking, and healthcare organizations, this book distills a decade of production experience. By the end, you'll be able to design, ship, and operate agentic AI systems with confidence. What you will learn

• Design agent systems with planner, router, and supervisor patterns
• Build production agents with the OpenAI Agents SDK
• Connect agents to enterprise tools and data through MCP
• Coordinate multi-agent workflows with A2A
• Engineer memory tiers that keep agents coherent and affordable
• Catch regressions with evaluation harnesses and quality gates
• Defend agents against prompt injection and memory poisoning
• Trace agent decisions, control token spend, and deploy safely

Who this book is for

This book is for AI engineers, application developers, solution architects, technical leads, and engineering managers who need to take AI agents beyond the prototype stage and into production. You should be comfortable reading Python and know your way around APIs, prompting, and tool calling. No prior agent framework experience is required.