Create a fully functional AI agent capable of reasoning, planning, and executing multiple sequential tasks!
Agents based on LLM are a new step in applied AI: they can reason, interact with tools, and achieve complex, multi-step goals. Thanks to new protocols such as MCP and A2A, agents can utilize external tools, access necessary data, and adapt to feedback. This book will guide you through the process of creating an AI agent from scratch — with clear and practical explanations, so you can build your own intelligent assistants!
In the book, you will learn how to:
- Implement the ReAct cycle (thought → action → observation)
- Use MCP to invoke tools in the agent's workflow
- Apply agent-based RAG for more relevant responses
- Create memory modules for facts, context, and goals
- Train agents in planning, reflection, and self-correction
- Develop specialized agents, including an agent for executing code
- Create systems of multiple interacting agents
Bestselling author Chongjun Hur and AI expert Yonhee Son will show you how an agent works “under the hood” — without hidden abstractions, black boxes, and dependency on specific frameworks. You will implement each component and gain a comprehensive understanding of how AI agents actually work.
About the Book
Build an AI Agent (From Scratch) is a step-by-step guide to creating a working AI agent: from the very basics to a full-fledged system for real-world use. You will connect the agent to tools, implement reasoning, and then expand it with memory, search, planning, reflection, and even inter-agent interaction. You can expect production-quality examples, ready prompts and configurations, as well as exercises to help improve the system. By the final chapter, you will have a fully operational AI agent that can be used as a foundation for your own projects.
