Directory · Claude Code extensions
Learn Claude Code
Tutorial that builds a minimal Claude Code-like agent harness in Python, step by step, for engineers learning agent internals.
★ 78,515 stars · last updated 2026-09-28 · Free, open source · View on GitHub
Best for
- Understanding how the Claude Code agent loop and tool calling work
- Learning harness mechanisms like subagents, context compaction, and task systems
- Building your own coding agent harness from scratch
- Readers who prefer runnable code alongside chapter READMEs
Our take
Educational repository rather than an extension you install into Claude Code; it consists of 17 root-level chapters (s01 to s17), each with a README in English, Chinese, and Japanese plus a runnable code.py. A legacy 12-lesson track remains in docs/ and agents/ during migration, and chapter numbers differ between tracks. The README is opinionated, arguing that agency comes from the model and the code's job is the harness.
Similar claude code extensions
- ECC: Plugin bundling agents, skills, hooks, memory, and security scanning to give Claude Code and other harnesses a structured dev workflow.
- Hermes Agent: Self-hosted, model-agnostic AI agent from Nous Research with persistent memory, self-improving skills, and chat-app access.
- REA (Reverse Engineer Anything): MCP server and CLI that lets coding agents inspect native binaries, JS/Electron apps, .NET, Android APKs and websites without source.
- Ponytail: A skill/plugin prompt that steers coding agents toward minimal code by reusing existing code, stdlib, and platform features first.
- CC Switch: Cross-platform desktop app for switching API providers and managing MCP, Skills, and Prompts across Claude Code, Codex, and other agents.
- UI UX Pro Max: Design intelligence skill for AI coding agents that generates tailored design systems, palettes, typography and UI patterns for apps.