Most explanations of AI agents stop at "call an LLM in a loop." This book goes further, walking you through building a fully functional agent from scratch — one capable of reasoning, planning, and carrying out multi-step tasks on its own.
Core Techniques You'll Implement
- The ReAct cycle: thought, action, observation
- Using MCP to call external tools from within the agent's workflow
- Agent-based RAG for pulling in more relevant context
- Memory modules that track facts, context, and goals
- Planning, reflection, and self-correction
Beyond the Basics
Later chapters extend the same foundation into specialized agents, including one that can execute code, and into systems where multiple agents interact with each other.
Why Build It From Scratch
Written by Jungjun Hur and Younghee Song, the book avoids hidden abstractions and framework lock-in, building every component from the ground up with production-quality examples, ready-to-use prompts, and hands-on exercises. By the final chapter, you have a working agent you can adapt into your own projects.
