SECTION

Foundations — Understanding Claude Code

From how Claude Code works to interactive usage, slash commands, tools, and the settings directory — the fundamentals that form the foundation of the agentic harness.

How It Works

The agentic loop, core components, and permission model of Claude Code, the agentic coding tool that runs in the terminal.

Features at a Glance

A hub page that lays out Claude Code's core features and extension layers on one page and guides you to each detailed document.

Interactive Mode

An at-a-glance guide to input methods, keyboard shortcuts, and permission modes in Claude Code's interactive REPL session.

Permissions and Plan Mode

The allow/ask/deny rules of the Claude Code permission system and the settings.json permissions block, the four permission modes, the subagent permission-mode inheritance rule (CC 2.1.219), and how Plan mode relates to the MoAI-ADK Implementation Kickoff Approval gate.

Slash Commands

Claude Code's slash commands — built-in commands, custom commands defined in markdown, scopes, plugin commands, and the use of core commands like /model, /agents, and /context at a beginner level.

Tools Reference

The purpose of Claude Code's built-in tools, the read/write distinction, permission rule formats, tool-selection best practices, and the subagent spawn semantics of the Agent tool — at beginner level.

.claude Directory

The structure of the .claude directory — the per-project configuration root Claude Code reads first every session — and the four setting scopes (project, user, enterprise, local), explained like explaining it to a friend.

This group covers the fundamentals you need before using Claude Code in earnest. You will learn, in order, how the agentic loop works, what features exist, how to type in interactive mode, how to use slash commands and tools, and where settings are stored. Everything you learn here — the loop, tools, permissions, the settings directory — is exactly the material MoAI-ADK builds its harness on.

Background reference
This page is background material on Claude Code itself, the platform MoAI-ADK runs on. MoAI-ADK’s own features are covered in the sections above it in the sidebar.

Info
Learning goal (one-line summary): Understand how Claude Code works and its core usage interfaces, so you can follow the workflow documents and MoAI-ADK harness design that come later without getting stuck.

Learning Flow

flowchart TD
    A[How It Works] --> B[Features at a Glance]
    B --> C[Interactive Mode]
    C --> D[Slash Commands]
    D --> E[Tools Reference]
    E --> F[.claude Directory]

First get the big picture with How It Works, then scan the feature map to see what tools exist. Next, learn the actual input methods with Interactive Mode and slash commands, and finish off behavior and environment with the Tools Reference and the settings directory — that completes your fundamentals.

Document	Description
How It Works	The agentic loop and core components
Features at a Glance	Full feature catalog and learning path
Interactive Mode	REPL, shortcuts, permission modes
Slash Commands	Built-in and custom commands, and the /moai relationship
Tools Reference	Built-in tools and permissions
.claude Directory	Settings directory structure and scopes
Permissions and Plan Mode	Allow/ask/deny rules, permission modes, plan mode

Once you have the fundamentals, move on to the next group, Context and Memory, to learn how to handle token costs — the starting point of MoAI-ADK tokenomics.