Overview

Claude Code by concept teaches Anthropic's official AI coding agent CLI through narrative explanation paired with annotated configuration and shell examples, organized into three progressive levels. Each level builds directly on the previous: Beginner establishes vocabulary and mental models for agentic tool use, Intermediate extends those models into custom workflows and integrations, and Advanced addresses production concerns, security, and architectural extensibility.

What Claude Code Is

Claude Code is not a chat interface you paste code into. It is a CLI process that runs on your machine, holds a live session with a Claude model, and calls tools — Read, Write, Edit, Bash, Glob, Grep, Agent — autonomously until a task is complete. That distinction shapes everything about how you learn and use it effectively.

%% Color Palette: Blue #0173B2, Orange #DE8F05, Teal #029E73, Purple #CC78BC, Brown #CA9161
%% All colors are color-blind friendly and meet WCAG AA contrast standards
 
graph TD
    USER["You<br/>natural language task"]:::blue
    CC["Claude Code CLI<br/>claude command"]:::orange
    MODEL["Claude Model<br/>Sonnet / Opus / Haiku"]:::teal
    TOOLS["Tools<br/>Read · Write · Edit · Bash<br/>Glob · Grep · Agent"]:::purple
    FS["Local Filesystem<br/>and Shell"]:::brown
 
    USER -->|"task description"| CC
    CC -->|"message + context"| MODEL
    MODEL -->|"tool call decision"| CC
    CC -->|"executes tool"| TOOLS
    TOOLS -->|"tool result"| CC
    CC -->|"tool result"| MODEL
    MODEL -->|"next tool call or final answer"| CC
    CC -->|"reads/writes/runs"| FS
 
    classDef blue fill:#0173B2,stroke:#000000,color:#FFFFFF,stroke-width:2px
    classDef orange fill:#DE8F05,stroke:#000000,color:#FFFFFF,stroke-width:2px
    classDef teal fill:#029E73,stroke:#000000,color:#FFFFFF,stroke-width:2px
    classDef purple fill:#CC78BC,stroke:#000000,color:#FFFFFF,stroke-width:2px
    classDef brown fill:#CA9161,stroke:#000000,color:#FFFFFF,stroke-width:2px

The diagram shows the core agentic loop: you give Claude Code a task, it reasons about which tool to call, executes that tool against your local filesystem or shell, feeds the result back to the model, and repeats until the task is finished or it needs more input from you. There is no copy-paste involved — Claude Code acts directly on your project.

Prerequisites

Before starting this track you need:

• Node.js 18 or later — Claude Code is distributed as an npm package
• An Anthropic API key (from console.anthropic.com) or an active Claude.ai Pro/Max subscription — Claude Code uses your account for model access
• Basic comfort with a terminal — you will run claude from the command line and read shell output

You do not need to know TypeScript, Python, or any specific language. Claude Code is language-agnostic.

Learning Path

%% Color Palette: Blue #0173B2, Orange #DE8F05, Teal #029E73, Purple #CC78BC, Brown #CA9161
%% All colors are color-blind friendly and meet WCAG AA contrast standards
 
graph TD
    B["Beginner<br/>16 sections<br/>Concepts 1–16"]:::blue
    I["Intermediate<br/>13 sections<br/>Concepts 17–29"]:::orange
    A["Advanced<br/>11 sections<br/>Concepts 30–40"]:::teal
 
    B -->|"internalize core concepts"| I
    I -->|"build custom workflows"| A
 
    B1["What is Claude Code?"]:::blue
    B2["Installation & Auth"]:::blue
    B3["Interactive Mode"]:::blue
    B4["Core Tools: Read, Write, Edit"]:::blue
    B5["Bash Tool"]:::blue
    B6["The Agentic Loop"]:::blue
 
    I1["Writing Effective CLAUDE.md"]:::orange
    I2["Hooks System"]:::orange
    I3["MCP Servers"]:::orange
    I4["Subagents"]:::orange
 
    A1["Advanced Hooks"]:::teal
    A2["Headless Mode & CI/CD"]:::teal
    A3["Security Model"]:::teal
 
    B --- B1
    B --- B2
    B --- B3
    B --- B4
    B --- B5
    B --- B6
 
    I --- I1
    I --- I2
    I --- I3
    I --- I4
 
    A --- A1
    A --- A2
    A --- A3
 
    classDef blue fill:#0173B2,stroke:#000000,color:#FFFFFF,stroke-width:2px
    classDef orange fill:#DE8F05,stroke:#000000,color:#FFFFFF,stroke-width:2px
    classDef teal fill:#029E73,stroke:#000000,color:#FFFFFF,stroke-width:2px

What Each Level Covers

Level	Coverage	Sections	Who It Is For
Beginner	0–40%	16	Anyone new to Claude Code or agentic CLI tools
Intermediate	40–75%	13	Engineers ready to customize workflows, hooks, MCP, and subagents
Advanced	75–95%	11	Engineers running Claude Code in CI/CD or extending it programmatically

Full Section Map

Beginner — 16 Sections

1. What is Claude Code? — Agentic CLI vs. AI chat, tool-calling vs. text generation, what "agent" means
2. Installation and Authentication — npm install -g @anthropic-ai/claude-code, API key vs. subscription auth, first run
3. Interactive Mode — TUI layout, prompt area, tool output, keyboard shortcuts
4. Core Tools: Read, Write, Edit — What each does, when Claude Code chooses each, file operation fundamentals
5. Bash Tool — Running shell commands, approval flow for dangerous commands, what it can execute
6. Glob and Grep — Codebase exploration, finding files by pattern, searching symbols in code
7. Your First Coding Task — Giving Claude Code a real task, watching the agentic loop, reading tool output
8. CLAUDE.md: Project Context — What it is, where it lives, what to put in it, scope hierarchy
9. The Agentic Loop — How Claude Code decides which tools to call, when it stops, multi-step reasoning
10. Slash Commands — /help, /clear, /compact, /cost, /model — what each does and when to use
11. One-Shot Mode — claude "task", non-interactive usage, scripting and CI pipelines
12. Model Selection — sonnet vs. opus vs. haiku, when to use each, /model command, /fast toggle
13. Cost Tracking — /cost command, token breakdown, managing spend on long sessions
14. Session Continuity — --continue, --resume, how sessions persist to disk
15. Safety and Permissions — What Claude Code can/cannot do by default, approval prompts
16. Plan Mode — /plan or EnterPlanMode, structured thinking before acting, approving the plan

Intermediate — 13 Sections

1. Writing Effective CLAUDE.md — What to include, @-file imports, anti-patterns, team governance
2. settings.json: Permissions — allowedTools, disallowedTools, additionalDirectories, auto-approve patterns
3. Hooks System — PreToolUse/PostToolUse/UserPromptSubmit/Stop, JSON stdin format, blocking tool calls
4. MCP Servers — What MCP is, connecting servers in settings.json, Playwright MCP, common use cases
5. Subagents (Agent Tool) — Spawning subagents, parallel vs. sequential, context isolation
6. Custom Slash Commands — Creating .claude/commands/*.md, $ARGUMENTS placeholder, team-shared commands
7. Skills System — .claude/skills//SKILL.md, inline vs. fork context modes
8. Memory System — /memory command, MEMORY.md, what to store vs. what to keep in CLAUDE.md
9. IDE Integration — VS Code extension, JetBrains plugin, inline suggestions, command palette
10. Context Management — /compact mechanics, token budget awareness, what gets compressed
11. Worktrees — claude --worktree <name>, parallel sessions, git worktree mechanics
12. Glob and Grep in Depth — Complex patterns, recursive search, combining tools for codebase archaeology
13. Parallel Subagent Patterns — Spawning multiple agents simultaneously, aggregating results, fan-out tasks

Advanced — 11 Sections

1. Advanced Hooks — Complex hook scripts, conditional blocking, modifying tool input, CI integration
2. Building Custom MCP Servers — MCP protocol, tool definitions, serving Claude Code new capabilities
3. Advanced CLAUDE.md Patterns — Conditional imports, role-specific instructions, team governance via CLAUDE.md
4. Subagent Orchestration at Scale — Orchestrator + specialist pattern, isolation modes, worktree subagents
5. Security Model — Permissions architecture, sensitive file handling, audit logging via hooks
6. Headless Mode and CI/CD — claude --headless, GitHub Actions integration, structured output
7. Performance Optimization — Selective context loading, targeted reads, reducing tool calls, prompt caching
8. Agent SDK — Building custom agents on the Claude Code SDK, tool registration, session management
9. Claude Code as MCP Server — Exposing Claude Code capabilities to other tools, integration patterns
10. Troubleshooting — Common errors, permission denied, context overflow, tool failures, diagnostics
11. Best Practices — CLAUDE.md governance, team conventions, cost optimization, task decomposition

How This Track Relates to By Example

The By Example track teaches Claude Code through short, densely annotated terminal sessions — you see exactly what command to type and what output to expect. This By Concept track teaches the same tool through narrative explanation and diagrams first, then annotated examples that reinforce the concept. Both tracks cover the same capabilities at equivalent depth; they differ only in pedagogical approach.

If you learn best by doing and then reading: start with By Example. If you learn best by understanding a model first and then practicing: start here.

How Each Section Is Structured

Every section in this track follows a consistent six-part format:

1. Concept title and introduction — what the concept is, why it matters, how it connects to what came before (2–3 sentences)
2. Diagram — a Mermaid flowchart or architecture diagram when the concept involves multiple components, data flows, or state transitions; omitted for trivial operations
3. Narrative explanation — how the concept works, when to use it, trade-offs, best practices, and pitfalls (3–10 paragraphs)
4. Annotated examples — shell sessions or configuration files with dense inline annotations using # => notation to document state, output, and reasoning at each step
5. Key Takeaway — the single most important insight from the section (1–2 sentences)
6. Why It Matters — how the concept connects to a real production concern such as cost, reliability, security, or team adoption (50–100 words)