# lintsinghua/claude-code-book

《御舆：解码 Agent Harness》42万字拆解 AI Agent 的Harness骨架与神经 —— Claude Code 架构深度剖析，15 章从对话循环到构建你自己的 Agent Harness。在线阅读网站：

Repository: https://github.com/lintsinghua/claude-code-book
Canonical: https://ross.abutalabs.com/products/claude-code-book
Homepage: https://lintsinghua.github.io
Language: Python
License Family: other
Topics: agent, agent-architecture, ai-agent, anthropic, book, chinese, claude-code, llm, mcp
Last push: 2026-08-17T06:10:12+00:00

## Health v2 (maintenance only)
Score: 59/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 98, release rhythm 35, longevity 11
- inputs: {"age_days": 155, "days_push": 16, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases, young, no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 4142, forks 816 (observed 2026-08-28T04:08:36.238399+00:00)

## What it is
A 420,000-word Chinese-language book (with English edition) that deeply analyzes the architecture of Claude Code as an AI Agent Harness, covering 15 chapters from the conversation loop to building your own agent harness. It is published as an open-source repository with an online reading site and 139 Mermaid architecture diagrams.

## Use cases
- understand how Claude Code's agent architecture works internally
- learn how agent tool systems and permission pipelines are designed
- study context compression and memory systems in AI agents
- build my own agent harness from scratch
- learn agent design patterns transferable to LangChain or AutoGen
- understand MCP integration and sub-agent scheduling
- deep dive into AI agent runtime architecture

## When to choose
- you want architectural analysis and design rationale rather than API usage tutorials
- you are building or designing a production-grade AI agent system
- you prefer learning through diagrams, design principles, and worked examples
- you read Chinese or English and want a systematic multi-week study path

## When to avoid
- you just want a quick-start guide or prompt engineering tips for using Claude Code
- you need official Anthropic documentation or authorized source code analysis
- you need a permissively licensed resource for commercial redistribution (CC BY-NC-SA)
- you want hands-on code libraries rather than a conceptual book

## Facets
- artifact type: learning-resource
- maturity: active
- function: documentation, agent-framework, mcp, llm-inference
- domain: large-language-models, tutorials, developer-tools, education
- platform: python
- tags: book, claude-code, agent-harness, architecture-analysis, chinese, online-reading, anthropic, ai-agents, web-server

## Member repositories
- lintsinghua/claude-code-book (main) score 59

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:08:36.238399+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T18:22:58.347975+00:00, confidence not recorded.
  - readme: https://github.com/lintsinghua/claude-code-book (fetched 2026-08-28T04:08:36.238399+00:00, sha 3f1fbd10ab3b)
  - homepage: https://lintsinghua.github.io (fetched 2026-08-29T09:13:57.805952+00:00, sha 41b12d4cf129)
- Data as of 2026-08-30T08:39:29.467469+00:00.