Preface
What the book covers and what it does not. Readers and how to read it.
Read the Preface
LLM Agent Design ArchitectureLLMエージェントの設計
Design agents around an LLM such as Claude so they remain usable.
Design agents around an LLM such as Claude so they remain usable.
The English title is LLM Agent Design Architecture. The Japanese title is LLMエージェントの設計.
The AI this book treats is primarily an LLM (large language model). That is the core of Claude. A plausible answer can be produced. There is no guarantee it matches the source text. Yesterday's conversation is gone unless you pass it in. This book is about assembling agents on that premise. It does not cover product how-tos.
Scope and terms are in the Preface.
A harness is the machinery for finishing today's task. This book is the design document for after the run. See Harness Engineering Mapping for the correspondence.
| Concern | Material | Role |
|---|---|---|
| Understand (origin of limits) | understanding-llm-through-claude-code | Why |
| Design | This book | What / How |
| Apply in operations | Management-of-software-systems-and-services | In preparation |
This is a record of knowledge from assembling agents. It is not official documentation from Anthropic or any other organisation. Comments and discussion are welcome in GitHub Issues.