This repository hosts a comprehensive technical book dissecting the architecture of Claude Code, an AI Agent Harness. It targets architects, senior engineers, and researchers seeking to understand the foundational design principles and engineering trade-offs behind production-grade AI agents, moving beyond basic usage guides. The book provides a transferable mental model and architectural blueprints applicable to various Agent frameworks, enabling readers to build or deeply comprehend complex agent systems.

How It Works

The book systematically deconstructs Claude Code's core components, including its dialogue loop, tool system, permission pipeline, context management, and memory architecture. It prioritizes explaining the "why" behind design choices, such as asynchronous generator patterns, layered configuration, context compression strategies, and multi-agent coordination. The analysis focuses on transferable design philosophies and engineering trade-offs, illustrated with 139 architectural diagrams, rather than serving as an API reference.

Quick Start & Requirements

The primary way to access the content is via the online reading website: https://lintsinghua.github.io/. The book itself has no installation requirements, but it analyzes a system potentially built with technologies like Bun, React/Ink, and Zod v4. A "how to read" guide suggests different engagement levels, with a full systematic study estimated at 2-3 weeks.

Highlighted Details

• Covers 15 chapters and 4 appendices detailing core subsystems: tool systems, permission pipelines, context compression, memory, hooks, sub-agents, and MCP integration.
• Features 139 Mermaid architectural diagrams and analyzes over 50 critical design decisions.
• Offers a transferable cognitive model applicable to frameworks like LangChain, AutoGen, CrewAI, or custom builds.
• Provides structured reading paths for quick overviews, deep dives, or reference lookups.

Maintenance & Community

The project welcomes community contributions via Issues and PRs for technical corrections and improvements. Links to external communities like "Linux.Do 社区" and "Star History" are present, but specific channels like Discord or Slack are not detailed in the provided text.

Licensing & Compatibility

The content is licensed under CC BY-NC-SA 4.0. This license prohibits commercial use and requires any derivative works to be shared under the same non-commercial, share-alike terms.

Limitations & Caveats

The book is an independent architectural analysis based on public information and product behavior, not affiliated with or endorsed by Anthropic PBC. The CC BY-NC-SA 4.0 license strictly limits commercial application. The repository provides access to the book's content, not a runnable codebase for the analyzed Claude Code system.