Start here

Steering a session

Teaching it your project

Permissions, safety and rules

More than one Claude

Outside the terminal

Making it real

Take one small repository from a bug nobody has found to a project Claude Code works in the way you want: a CLAUDE.md it follows, a skill it repeats, a hook it cannot argue with, subagents that keep your context clean, and a review that runs without you. Every terminal block on this site was captured from a real session.

Claude Code

Everything in this course, and where it is going

What you need

Claude Code course: a complete, practical tutorial

What you are going to build

It can teach you the tool

Install Claude Code and run your first prompt

Install it and ask it something

One turn, as it really ran

Harness, not model

How the Claude Code agent loop works

What it actually is

The same thing that helps people

How Claude Code searches a codebase

How it finds things

Permission is not undo

Claude Code permissions: what the prompt is asking

The permission prompt

In the session

A plan is a proposal

Claude Code plan mode: propose before editing

Plan before it writes

Commands are not covered

Claude Code checkpoints and the rewind menu

Undo: rewind and checkpoints

What fills it, roughly in order

When compaction gives up

Claude Code context window and compaction explained

The context window, and compaction

When it keeps missing

How to interrupt and steer Claude Code mid-task

Interrupting and correcting

Headless runs are hidden

Claude Code sessions: resume, continue and fork

Sessions: resume, continue, fork

It is a strong suggestion

Writing a CLAUDE.md Claude Code actually follows

CLAUDE.md that gets followed

Rules that fight

Claude Code memory: project, user and org scopes

Where memory lives

Seeing what exists

Custom slash commands in Claude Code

Slash commands you write yourself

Short bodies

Claude Code skills: teach it a repeatable procedure

Skills: a procedure it can repeat

default

Or ask for it

Claude Code statusline: show context, cost and branch

The statusline, and output styles

A reasonable default day

It is on screen

Claude Code permission modes: manual, accept edits, plan, auto

Permission modes

Start with deny

Claude Code permission rules in settings.json

settings.json: allow, ask, deny

What it is not

The Claude Code Bash sandbox explained

Sandboxing

What the program gets, and what it says back

The other way to block

Claude Code hooks: block a tool call with your own code

Hooks: a rule that cannot be argued with

Why hooks are the answer

Write a Claude Code PreToolUse hook in Python

Writing a hook that blocks a bad edit

When to delegate

Claude Code subagents: keep your context clean

Subagents

What does not come with you

Run parallel Claude Code sessions with git worktrees

Worktrees: real parallel work

One at a time first

Claude Code agent view: manage background sessions

Background sessions and agent view

Watching instead of waiting

Claude Code headless mode: claude -p and JSON output

Headless mode and JSON output

Unattended means the rules matter

Automated code review with Claude Code and GitHub Actions

Review, and Claude in your pull requests

It is code you are running

Connect Claude Code to external tools with MCP

MCP: tools you did not write

Conventions, review skill and protected-file hook for this project

link-shortener-kit

What good looks like

Claude Code plugins: share skills, agents and hooks

Plugins: packaging a setup

The rules are the supervision

Claude Code routines, scheduled tasks and remote control

Work that runs without you

Measure first

Claude Code in a monorepo or large codebase

Working in a large codebase

When speed is worth it

Claude Code models and managing cost

Models, cost and fast mode

The finished repository, and the lesson each piece came from

Where to go next

Claude Code: a complete project setup, end to end

Set up a repository end to end

Start here quiz

What does Claude Code do before answering a question about your project?