Preface

Building Agents with OpenAI Agents SDK comes at a time when there is a major shift in AI and how we work. The first wave came with ChatGPT, a large language models (LLM) in a consumer application, which gave everyone powerful new ways to generate and work with natural language. Now, we are moving into the next phase: AI agents. Unlike traditional software, which is built around rigid, deterministic rules, such as “if X, then Y,” agents can handle ambiguity, can adapt to changing goals, and can reason through steps to accomplish tasks. They are not limited to answering questions; they can plan, orchestrate workflows, retrieve and synthesize information, use external tools, and even collaborate with one another. In doing so, they are beginning to take on work that once seemed possible only for humans. This transformation is still in its early days.

OpenAI Agents SDK is at the center of this change. It provides a practical and the best framework for building, deploying, and managing agents, from simple assistants to complex multi-agent systems. Just as Microsoft PowerPoint is a platform to build slide deck presentations, OpenAI Agents SDK is poised to become the way developers and tech enthusiasts build AI agents and agentic systems.

This book is your guide to mastering that skill. It takes a hands-on, example-based approach. We start in the beginning by learning the anatomy of an agent, its strengths and weaknesses, and how OpenAI Agents SDK works. We then deep dive into each of the SDK’s capabilities, such as tools, memory, knowledge, multi-agent handoffs, and guardrails. We then take everything we’ve learned and build a full end-to-end practical AI agent solution.

By the end, you won’t just understand AI agents in theory; you’ll have built full, end-to-end agentic systems with OpenAI Agents SDK. You’ll be equipped not only to answer the question, “What is an AI agent?”, but also to design and implement AI agents in practice. The best way to learn is by building, and this book is written to get you building AI agents.

The book is structured into three parts:

• Part 1, AI Agents, lays the conceptual and practical groundwork for the rest of the book. We start by defining what an agent is, how it differs from a traditional chatbot or API client, and why agents have become a compelling pattern for automating work across software, data, and business processes. We will then build a mental model of an AI agent that you can use throughout the book: the brain of an agent (the model), the hands and eyes of an agent (the tooling interface), and the reference textbook (memory and knowledge), all under explicit constraints and guardrails. We will then learn about the core primitives of the best framework to build AI agents – OpenAI Agents SDK. 
• Part 2, OpenAI Agents SDK, is the technical core of the book. Here, we move from “agent as an idea” to “agent as an engineered system,” using OpenAI Agents SDK as our primary vehicle. Each chapter deepens one specific capability and the primitives set of OpenAI Agents SDK. For each component, we discuss the theory and then apply it to examples and demos as we build practical AI agents and agentic systems. Specifically, we will extend your agent with tools (from simple functions to hosted tools and agents-as-tools), add memory and knowledge retrieval, compose multi-agent systems with deliberate handoffs, and learn how to manage agentic systems.
• Part 3, Build AI Agents, is the key part of the book. Here, we will assemble an end-to-end agentic system using OpenAI Agents SDK, not as isolated features but as a coherent application. We will take the components introduced earlier (tools, memory and retrieval, sessions, model and context management, guardrails, multi-agent orchestration, etc.) and integrate them into production-shaped applications and workflows with clear inputs and outputs.