Discover Claude Code and install it

Why Claude Code changes the game

Most AI coding tools just complete a line. You type the start of a function, the tool guesses the rest. That's useful, but it's passive assistance: all the orchestration intelligence — what to do, in what order, how to verify — stays in your head. Claude Code does something different: it's an agent. It reads your files, runs commands in your terminal, writes code, launches tests, and reasons across multiple steps to reach a goal you give it in plain language.

The difference is fundamental. A classic chatbot replies with text you have to copy-paste yourself. An autocomplete suggests the next line. Claude Code acts: if you ask it to "fix the bug in the publishing script," it will find the relevant file, read the code, understand the error, propose a fix, apply it, then rerun the script to confirm it works. You supervise. It's the shift from being the doer to being the director.

For Lea, that means she won't be "asking an AI for snippets of text." She'll hand it a complete mission — "publish this topic across my 4 networks" — and Claude orchestrates everything: research, writing, visuals, publishing. And as you'll discover, what makes this approach truly powerful is that it compounds: every automation you build stays available forever.

The agentic loop: read, execute, write, reason

Under the hood, Claude Code runs in a loop: it observes the state of your project, decides on an action, executes it, observes the result, then decides what comes next. The loop continues until the goal is reached — or until it needs you to make a call. It's exactly like delegating to a colleague: you give the mission and the constraints, they handle the intermediate steps.

flowchart LR
  U["You: mission in plain language"] --> C["Claude Code"]
  C --> L["Reads your files"]
  C --> X["Runs commands"]
  C --> W["Writes code"]
  L --> R["Reasons and chains the steps"]
  X --> R
  W --> R
  R --> OK["Goal reached"]

This loop has an important practical consequence: the quality of what you get depends on the quality of the mission you give. "Write me a post" will produce something mediocre. "Write a 150-word LinkedIn post about zero waste, warm tone, ending with an open question" will produce something usable. This whole course teaches you to encode those specifics once and for all, instead of retyping them every time.

Installing Claude Code

Claude Code installs via npm, the Node.js package manager. So you need Node.js 18 or later — if you don't have it, download the LTS version from the official Node.js site; installation takes two minutes. Then open a terminal and run:

# Check that Node.js is installed (18+ required)
node --version

# Install Claude Code globally
npm install -g @anthropic-ai/claude-code

# Move into your project folder
cd my-project

# Launch Claude Code
claude

On first launch, Claude Code asks you to sign in to your Anthropic account. Two options: a Claude subscription (Pro or Max), or an API key from the Anthropic console billed by usage. To get started, the subscription is simpler: a predictable monthly quota, no surprises. Once authenticated, you're ready.

On Windows, Claude Code runs natively in PowerShell, but many users prefer to launch it inside WSL (Windows Subsystem for Linux) or Git Bash to get a more standard Unix command environment. If a shell command behaves strangely, that's the first thing to check.

Your first session

Once inside a session, you talk to Claude in plain language, like a colleague. But a few built-in commands — the ones starting with / — are worth knowing from day one:

• /help — lists all available commands.
• /init — analyzes your project and generates a first CLAUDE.md file (the project memory; we dedicate chapter 8 to it).
• /clear — wipes the conversation to start from a clean context.
• /model — shows and changes the model in use.
• Esc — interrupts Claude mid-action if you see it heading in the wrong direction.
• @path/to/file — references a specific file in your message so Claude reads it.

Take five minutes to play: ask it to "list the files in this folder and describe what you understand about the project." You'll see the mechanics live: it proposes running a command, asks for your permission, runs it, reads the result, and gives you a summary. That systematic permission prompt can feel heavy — which is exactly the subject of chapter 2.

VS Code extension or terminal?

Claude Code works two ways inside VS Code: the official extension (sidebar interface) or the integrated terminal (the claude command). Both run exactly the same engine — the difference is purely ergonomic.

Which model should you use?

Claude Code lets you pick the model via /model. The simple rule: use the most capable model available for almost everything — technical reasoning, building skills, API integrations, debugging. These tasks require sustaining a chain of reasoning over many steps, and that's exactly where the most powerful models pull ahead.

Reserve the lighter (and cheaper) models for simple, repetitive tasks: reformatting a file, renaming variables, generating variants of text that's already approved. For this course, keep the most powerful model everywhere — orchestration and subagent coordination benefit from it directly. One reasoning error in the middle of a 10-step chain costs far more in time than the price difference between two models.

What you'll build in this course

Each chapter adds a brick to Lea's system. Chapter 2: the settings that make work flow. Chapter 3: a first /post skill that publishes to one network. Chapter 4: the brand voice so the posts sound human. Chapter 5: a hook that blocks any non-compliant publication. Chapter 6: subagents that publish to 4 networks in parallel. Chapter 7: the capstone command /plan-week that schedules an entire week. Chapter 8: the project memory that makes everything survive from one session to the next.

The order isn't arbitrary: it's the natural progression of any serious automation. First make it work once (skill), then make it reliable (hook), then scale it (subagents), then capitalize on it (memory). You'll be able to replay this exact sequence on any other project — accounting, customer support, competitive intelligence.