Claude Code — The Power Tool

Claude Code is Claude, but on your computer instead of in a browser tab.

When you use Claude on claude.ai, you're typing in a web page. Claude can only see what you paste into the conversation. It's helpful, but it's like talking to someone over the phone — you have to describe everything.

Claude Code changes this. It runs in your terminal (we'll explain what that is in a moment) and can directly see your files, read your documents, and make changes to things on your computer. It's like inviting that capable colleague from Module 1 to sit at your desk instead of calling them from another room.

What this means in practice:

• "Read through this folder of documents and summarise the key points" — Claude Code opens and reads each file itself. You don't paste anything.
• "Organise these 50 files into folders by topic" — Claude Code actually moves the files. You don't do it manually.
• "Find every mention of 'budget' across my notes and compile them" — Claude Code searches, finds, and compiles. You get the result.
• "Edit my report — fix the grammar and tighten the writing" — Claude Code opens the file, makes changes, and shows you what it did.

If you're a developer, it can also write and edit code, run tests, manage git repositories, and handle complex programming tasks. But you absolutely don't need to be a developer to get value from it. Anyone who works with files and documents on their computer can benefit.

Claude Code runs in something called the terminal (also called the command line). If you've never used it, it looks like a black screen with white text, and it can feel like something only programmers touch.

Let's demystify it. The terminal is just another way to tell your computer what to do. Instead of clicking icons and menus, you type short commands. That's the entire difference. Your computer's the same. Your files are the same. You're just talking to it differently.

Think of it this way: using your regular desktop is like pointing at things in a restaurant and saying "I'll have that." Using the terminal is like saying the name of the dish. Same restaurant, same food, different way of ordering.

You'll need to type about five commands total to get Claude Code installed and running. We'll walk through each one. After that, you mostly just talk to Claude in plain English — the terminal is just where the conversation happens.

How to open the terminal

On Mac: Press Cmd + Space, type "Terminal", press Enter. That's it.

On Windows: Press the Windows key, type "PowerShell", press Enter. (If you have Windows Terminal installed, use that instead — it's nicer to look at.)

On Linux: You probably already know. But: Ctrl + Alt + T on most distributions.

You'll see a window with a blinking cursor. This is where we'll work.

Before we install Claude Code, we need one thing on the computer first: Node.js. Think of Node.js as the engine that Claude Code runs on. You install it once and forget about it.

Step 1: Install Node.js

Go to nodejs.org. Download the version labelled "LTS" (the one on the left — it stands for Long Term Support, which just means "the stable one"). Run the installer. Click through the defaults. Done.

To verify it worked, open your terminal and type:

node --version

You should see a version number like v22.x.x. If you do, Node.js is installed. If you get an error, close the terminal, reopen it, and try again — sometimes it needs a fresh start to recognise new installations.

Step 2: Install Claude Code

In your terminal, type:

npm install -g @anthropic-ai/claude-code

This tells your computer to download and install Claude Code. It might take a minute. You'll see some text scrolling by — that's normal. Wait for it to finish.

What just happened: npm is a tool that came with Node.js. It downloads and installs packages (think of them as apps for the terminal). We just used it to install Claude Code globally on your computer (that's what the -g means — available everywhere, not just in one folder).

Step 3: Start Claude Code

Navigate to a folder you want to work in. For your first time, let's use your Documents folder:

Mac/Linux:

cd ~/Documents

Windows:

cd C:\Users\YourName\Documents

(Replace "YourName" with your actual Windows username.)

Now start Claude Code:

claude

That's it. One word. Claude Code will start up and connect to your Claude account. The first time, it'll ask you to log in through your browser — follow the link, authorize it, and come back to the terminal.

You should see a prompt where you can type messages to Claude, just like on the website — except now Claude can see your files.

Claude Code is running. Let's do something useful.

Type this (or something like it, adapted to whatever's actually in your Documents folder):

What files are in this folder? Give me a quick summary of what's here.

Claude will list your files and describe what it finds. It's actually looking at your computer, not guessing.

Now try something more useful:

Find the most recently modified document in this folder and give me a one-paragraph summary of what it's about.

Claude opens the file, reads it, and summarises it. You didn't paste anything, didn't upload anything, didn't copy-paste text into a browser. Claude just... read your file.

A few more things to try in your first session:

Ask Claude to organise something:

Look at the files in this folder. Suggest a way to organise them into subfolders by topic. Don't move anything yet — just show me the plan.

(Notice the "don't move anything yet" — Claude Code can make changes, so it's good practice to ask for a plan first, then approve it.)

Ask Claude to help with a document:

Open [filename] and review it. Point out anything that's unclear or could be written more concisely. Show me the specific changes you'd suggest.

Ask Claude something about your files:

Search through all the documents here and find anything related to [topic]. List what you found with a one-line summary of each.

The permission system

When Claude Code wants to do something significant — create a file, delete something, run a command — it asks you first. You'll see a prompt like:

Claude wants to: Create file "summary.md"
Allow? (y/n)

This is a safety net. Claude won't silently modify things without your approval. Read what it's proposing, and approve if it looks right. If it doesn't look right, say no and explain what you wanted instead.

As you get comfortable, you can adjust these permissions to let Claude handle more things automatically. But for now, the default "ask before acting" behaviour is exactly right.

On claude.ai, we set up Custom Instructions. Claude Code has its own version: a file called CLAUDE.md.

It's a plain text file that lives in your home folder (or any project folder). When Claude Code starts, it reads this file and uses it as context. Think of it as a briefing document that's always on Claude's desk.

What to put in it

Here's a simple starter CLAUDE.md:

# About Me
I'm [your role/situation]. I mainly use Claude Code for [your primary use cases].

# Preferences
- Keep responses concise — explain what you're doing, but don't over-explain
- Use British English
- When suggesting file changes, show me the plan before making changes
- Currency: use ₹, not $

# My Setup
- Operating system: [Mac/Windows/Linux]
- Main folders I work in: [list them]

You don't need more than this to start. As you use Claude Code more, you'll naturally discover preferences you want to add. The file grows with you.

Where to put it

Create a file called CLAUDE.md in your home directory:

Mac/Linux: ~/.claude/CLAUDE.md Windows: C:\Users\YourName\.claude\CLAUDE.md

You can create it with any text editor — Notes, TextEdit, VS Code, whatever you have. Save it, and Claude Code will automatically read it next time it starts.

Why this matters for tokens

Without CLAUDE.md, you spend the first few messages of every session re-explaining your preferences. "Use British English." "Keep it concise." "I'm working on marketing stuff." Every time.

With CLAUDE.md, that context loads automatically. Zero extra tokens from you, and Claude starts every session already knowing how you work.

Claude Code has various settings you can adjust. Let's be honest about which ones matter right now and which ones you can ignore.

Worth setting up now

Your editor preference. If Claude Code needs to show you a file, which editor should it open? Set this if you have a preferred text editor.

Permission levels. The default "ask before doing anything" is good for beginners. As you get comfortable, you might want to let Claude handle routine tasks (like reading files) without asking. You can adjust this over time.

Model selection. If you're on Pro, Claude Code uses your Pro subscription. You don't need API keys. It'll use the best model available to you.

Ignore for now

Hooks, MCP servers, custom agents. These are powerful but you don't need them yet. They're for people who want to extend Claude Code with custom capabilities. The later modules cover each of these in depth. But they're genuinely optional.

Advanced configuration files. Claude Code can be extensively customised through various config files. The CLAUDE.md we set up above handles 90% of what you need. The rest can wait.

The general principle: start simple, add complexity only when you feel a specific need for it. Claude Code works well out of the box.

You'll come across these terms as you explore Claude Code. Let's define them simply so they stop being mysterious. You don't need to set any of these up right now.

Skills are pre-written instructions that activate when Claude encounters specific situations. Think of them as specialised knowledge Claude can tap into. A "git commit" skill, for example, tells Claude exactly how to create clean, well-formatted git commits. You don't write the instructions every time — the skill handles it.

Agents are like sub-tasks. When Claude Code encounters something complex, it can spin up a smaller, focused "agent" to handle part of the work — like sending a research assistant to look something up while the main conversation continues.

Hooks are automated triggers. "Every time Claude creates a commit, run this check." "When a session starts, sync these files." They're for automation, and they're entirely optional.

The honest take: These features are what make Claude Code extremely powerful for advanced users. But they're also what makes it feel overwhelming when you read about it online. The good news is you can completely ignore all three and still get massive value from Claude Code. They're layer-two features. Get comfortable with the basics first.

You can stop here if you'd like. You now know what Claude Code is, have it installed, and had your first session. That's a solid working foundation.

Module 4 takes this further — it's the setup that makes Claude Code genuinely token-efficient and tailored to your workflow. That module wraps up the free portion of this guide.

Take a break if you need one. Make some chai. Module 4 will be here when you're back.