ClawdCode Guide

A step-by-step tutorial on building a production-grade CLI coding agent.

The Formula

Coding Agent = LLM + System Prompt + Context + Tools

Component	Role
LLM	Reasoning engine
System Prompt	Identity, constraints, style
Context	Project info, conversation history
Tools	File I/O, shell, search — the agent's hands

Structure

Getting Started

#	Topic	What you'll learn
01	Coding Agent Overview	Core concepts, design philosophy
02	Project Setup	Tech stack, Hello World agent
03	CLI Entry	yargs, middleware, version check

Core Architecture

#	Topic	What you'll learn
04	Agent Core	Agent class, Agentic Loop
05	System Prompt	Prompt architecture, Plan mode
06	Tool System	Tool abstraction, built-in tools
07	Execution Pipeline	7-stage pipeline, permission model

System Integration

#	Topic	What you'll learn
08	Context Management	Token counting, auto-compaction
09	UI System	Ink (React for CLI), Markdown
10	MCP Protocol	External tool discovery, protocol
11	State Management	Zustand store, session persistence
12	Command History & Queue	Command history, queue system

Extensions

#	Topic	What you'll learn
13	Slash Commands	Command system, custom commands
14	Interactive Commands	Model/theme interactive selection
15	Streaming & Themes	Streaming output, theme persistence
16	Skills System	Agent skill modules
17	Hooks System	Lifecycle hooks

Tech Stack

TypeScript · Bun · Ink · Zustand · OpenAI SDK · Zod · MCP

Requirements

• Node.js >= 18
• Bun >= 1.0
• An OpenAI-compatible API key

Install

npm

npm install -g clawdcode

bun

bun add -g clawdcode

source

git clone https://github.com/kkkhs/ClawdCode.git
cd ClawdCode && bun install && bun run dev

Run

clawdcode                          # interactive mode
clawdcode "analyze this project"   # with initial message
clawdcode --continue               # resume last session

Resources

• OpenAI Function Calling
• Ink
• Model Context Protocol

---

Ready? Start with 01. Coding Agent Overview.