An educational, course-style web application that breaks down the architecture and internals of Claude Code into structured, readable modules.
This project was created after the Claude Code source map became publicly available. The goal is to provide a persistent, community-accessible knowledge base for understanding how Claude Code works under the hood — independent of any DMCA takedowns or content removal that may affect the original source.
The app is organized into 9 guided modules covering the full Claude Code architecture:
| # | Module | Topic |
|---|---|---|
| 0 | What Is Claude Code? | Orientation, mental model, tech stack, repository map |
| 1 | Bootstrap & Startup | From binary to REPL in under 200ms |
| 2 | The Core Agent Loop | The while(true) that drives everything |
| 3 | The Tool System | How the 45+ built-in tools are defined and executed |
| 4 | Permissions & Safety | Permission modes, rule evaluation, bash classifier, sandbox |
| 5 | Context Management | System prompt assembly, compaction strategies, memory |
| 6 | The Terminal UI | React + custom Ink fork rendering to ANSI escape sequences |
| 7 | Extensibility | Hooks, MCP, Skills, Plugins |
| 8 | Multi-Agent Orchestration | Subagent spawning, worktree isolation, task system |
Requires Node.js (v18+).
npm install
npm run devOpen the URL printed in the terminal.
This project was vibe-coded in a single session with the primary purpose of organizing and preserving knowledge about the Claude Code architecture. It is provided as-is. Expect rough edges, incomplete sections, and bugs. The content is the focus, not the polish.
Contributions and pull requests are welcome.
This project is licensed under the MIT License. See LICENSE.md for details.
