chemark/ai-agent-book

What it solves

Building a reliable AI agent is harder than it looks. An agent must combine a large language model (LLM), the right context, and a set of tools — and the engineering details matter a lot. This book gives you a complete, hands-on path to understand and build agents, covering everything from basic concepts to advanced collaboration and self-evolution, so you can move from theory to working code without guessing.

How it works

The book is built around one core formula: Agent = LLM + context + tools. It is organized into 10 chapters, each with a focused theme and runnable code. The chapters progress from fundamentals (Chapter 1) through context engineering (Chapter 2), user memory and knowledge bases (Chapter 3), tools and MCP (Chapter 4), coding agents (Chapter 5), evaluation (Chapter 6), model post-training (Chapter 7), agent self-evolution (Chapter 8), multimodal and real-time interaction (Chapter 9), and finally multi-agent collaboration (Chapter 10). Throughout, you get a mix of theory, practical guidance, and 92 hands-on experiments. Some experiments require cloning external repositories, and the README provides one-click clone commands, plus a list of suitable API key providers (like DeepSeek or Kimi) to get started.

Who it’s for

This is for developers, students, and researchers who want to learn AI agents by doing. It suits people with some coding background who are willing to run experiments and follow a structured learning path. The book includes recommendations on how to read efficiently, and each chapter’s experiments are clearly labeled (runnable, reproduce, or design-only), so you can pick your level of depth.

Highlights

• Core formula: Agent = LLM + context + tools
• 92 open-source experiments across 10 chapters
• Covers context engineering, memory, tools, coding agents, evaluation, post-training, self-evolution, multimodal, and multi-agent systems
• Includes a one-click script to clone 19 external repositories needed for deeper experiments
• Provides learning suggestions and API key recommendations for easy setup