What it does

This repo is a book-length study of Claude Code’s source code — not the code itself, but 34 chapters of annotated walkthroughs. Each chapter maps to a specific file or function, cites line numbers, and extracts reusable design patterns. The author organizes everything around the runtime lifecycle: cold start, query loop, tool registry, agent orchestration, network layer, terminal UI, and memory subsystem.

The interesting bit

Most “learn from source code” projects settle for high-level diagrams. This one goes file-by-file through a real commercial product, including the unglamorous bits: 11 configuration migration files, a 35-line custom store for cross-process state, and a feature-flag system that compiles two products from one codebase. The author even documents the “Buddy pet” — a randomly generated animal that sits next to the prompt input.

Key highlights

• 34 chapters with exact source line references and code snippets
• Coverage of System Prompt engineering, context compression, and Prompt Cache cross-cutting
• Tool system with three taxonomies: family, runtime leaf, and feature-gated
• Bridge IPC for remote control from phone/Web/Desktop to local CLI
• Custom Ink/React reconciler for terminal UI, plus Vim mode and voice input
• Six appendices with lookup tables for tools, commands, hooks, agents, and task types
• Multiple reading paths: 7-chapter intro, 9-chapter AI engineering deep-dive, or full 34-chapter marathon

Caveats

• The analysis itself is in Chinese; English readers get a shorter README-en.md but the full chapters are untranslated
• The underlying Claude Code source is proprietary; this is third-party interpretation, not official documentation
• Some claims (e.g., “best AI Coding product”) are the author’s assessment, not an independent benchmark

Verdict

Worth bookmarking if you’re building AI agents and want to see how a polished product handles the boring parts: config migrations, permission callbacks, and terminal layout engines. Skip it if you need runnable code or official Anthropic docs.