Claude Code
From Zero to Dangerous

Module 1What You're Building

By the end of this guide, you will have a working AI development environment that lets you:

• Talk to Claude in your code editor - not a browser tab, your actual workspace
• Give Claude persistent memory - it remembers your business, your clients, your preferences across sessions
• Connect APIs - Google Drive, Sheets, Gmail, GitHub, and more talk to Claude directly
• Build reusable workflows - one prompt does what used to take 30 minutes of manual work
• Version control everything - nothing is ever lost, every change is tracked
• Deploy live websites - go from a file on your laptop to a live URL in minutes

How It All Fits Together

The Outcome

You go from "I use ChatGPT sometimes" to "I have an AI operating system that knows my business." This is not about coding. It is about leverage.

Claude Pricing

Claude Code requires a paid Anthropic subscription. Here are your options:

Tool Alternatives

Claude Code runs inside a code editor. You have several options, but the choice is straightforward - pick the one with the best native integration and the lowest barrier to entry.

Prerequisites

You only need five things to get started, and you probably already have most of them.

Module 1.5 Your First Week: How to Actually Learn This

The tools in this guide would have taken a team of developers months to set up five years ago. Now you can do it in an afternoon. But that speed creates a trap: you can install everything without understanding anything. This module is your antidote.

The Foundation 30 - Don't Let Setup Become the Project

VS Code, GitHub, and Cloudflare should take about 30 minutes to get through, not an afternoon. If you're past the hour mark on any single one, stop, type your blocker into Claude as a question, and keep moving. Setup is supposed to be fast. The value is in what you build on top of it.

Week 1 Rule: Watch Claude Work, Don't Just Skim the Output

Most people start issuing commands and skimming results. That's fine once you know what you're doing. In week one, slow down and actually read what Claude produces. Especially the thinking - if Claude explains WHY it's doing something, that's a free lesson.

• When Claude uses a term you don't know - CDN (content delivery network), MCP (model context protocol), repo (repository), commit (a saved snapshot), staging (a test environment), DNS (the internet's address book), CLI (command-line interface), agent (an AI given tools and autonomy) - stop. Ask: "What does [term] mean? Explain it like I've never coded." Don't move on until you have a plain-English answer.
• Paste that answer into a glossary.md file in your workspace. Write it in your own words - that's the test of whether you actually understood it.
• When something clicks, write it down. "I now understand why version control matters because..." That's the learning. Don't just feel it, capture it.

Build Your Glossary From Day One

Create a file called glossary.md in your workspace root. Every time you hit an unfamiliar term, add it. Write a 3-5 sentence paragraph for each - something you can re-read in 30 seconds when the term comes back. The goal isn't to memorise. It's to build a reference doc that compounds.

Who Owns What - The Control Plane

One of the most confusing things early on is understanding which service controls what. Here's the map. Once you have this, you'll never be confused about "where to go to fix X".

Module 2VS Code Setup + Essential Shortcuts

VS Code is your home base. It is where you will talk to Claude, edit files, run commands, and manage your projects. It is free, runs on everything, and has the best Claude Code integration.

Install + First Launch

1. Download VS Code from code.visualstudio.com
2. Install it (drag to Applications on Mac, run the installer on Windows)
3. Open it. You will see a Welcome tab - close it.
4. Open a folder: Cmd+O (Mac) or Ctrl+O (Windows) → create a new folder called my-workspace on your Desktop → open it

Windows, Folders, and Context

This is the single most confusing thing for new users. Understanding it will save you a lot of frustration.

The folder IS the context

When you open a folder in VS Code, Claude Code loads everything scoped to that folder: custom skills, rules, environment variables, and project memory. Without an open folder, Claude is running in a blank context. It still works, but it has no project-specific knowledge.

Example: imagine you have a folder called michaels-business on your Desktop. Inside it you have built custom Claude skills, added rules, and stored client notes. When VS Code has michaels-business open as the workspace, Claude can see all of it. If you close that folder or open a different one, Claude loses that context immediately.

The right way: open a new window first

If you want to open a new project folder without losing your current tabs, always open a new window first:

Then use File → Open Folder in that new window. Your existing window and all its tabs stay intact.

One folder per project type

A good habit: each VS Code window has one folder open, and that folder matches the kind of work you are doing. If you have three different projects, run three windows, each with the right folder. Tabs within a window all share the same folder context.

Think of it this way:

• Window 1: michaels-business open → Claude knows everything about Michael's project
• Window 2: personal-site open → Claude knows about your personal site
• Window 3: no folder open → Claude has no project context (generic mode)

What to do if you have tabs open with no folder

If you find yourself with lots of tabs open but no folder loaded, this is a common situation. Your options:

1. Open a new window (Cmd+Shift+N / Ctrl+Shift+N), open the correct folder there, and start new conversations in that window
2. If you need to continue an existing conversation from the no-folder window, copy the context out of that chat, then paste it into a new conversation in the correct window
3. Accept that the no-folder tabs are running without project context, which is fine for general questions but not for project-specific work

Install Node.js (Required for Claude Code)

Essential Extensions

Open the Extensions panel: Cmd+Shift+X (Mac) or Ctrl+Shift+X (Windows). Search and install these:

With extensions installed, VS Code transforms from a blank editor into a purpose-built AI workspace. Next, learn the handful of keyboard shortcuts that will save you the most time.

VS Code Shortcuts That Matter

You do not need to memorise 200 shortcuts. These 12 cover 90% of daily use:

Laptop Power Shortcuts

These are not VS Code shortcuts - they are operating system shortcuts that make you faster at everything. Most people never learn them:

Workspace Folder Structure

Here is the recommended folder structure for your workspace. You do not need to create all of these right now - Claude will create them as needed. But understanding the layout helps:

Why this matters: Claude Code reads and writes files in your workspace. A clean structure means Claude always knows where to find things and where to save outputs. It also means you can find things yourself without digging through chaos.

Module 3Git + GitHub - Your Safety Net

Git tracks every change you make. GitHub stores it in the cloud. Together, they mean you can never lose your work and you can always undo a mistake. Think of it as "unlimited undo for your entire project."

Why This Matters for Non-Coders

• Nothing is ever lost - deleted a file by accident? git checkout brings it back.
• Claude uses git automatically - Claude Code commits its work so you have a trail of everything it built.
• Collaboration - share your workspace with a partner, developer, or another AI tool.
• Free hosting - GitHub Pages gives you free live URLs for any HTML file you push.

Setup Steps

1. Create a free account at github.com
2. Install the GitHub CLI (handles authentication cleanly):

3. Create a new repository on GitHub (green "New" button). Name it my-workspace. Make it Private.
4. Open VS Code terminal (Ctrl+`) and run these commands one at a time:

Immediately after git init, create your .gitignore file so sensitive files are never tracked:

Now make your first commit and push:

Daily Git Workflow (3 Commands)

You do not need to learn complex git. These 3 commands cover 95% of daily use:

Your First Live URL - GitHub Pages

Once your code is pushed to GitHub, you can make it publicly accessible in a few clicks. This gives you a real https:// URL that anyone can visit.

1. Push your HTML file to your GitHub repo (you just learned how)
2. Go to your repo on github.com
3. Click Settings → Pages (left sidebar)
4. Under "Source", select main branch and / (root) folder → click Save
5. Wait 1-2 minutes. Your site is live at https://YOUR-USERNAME.github.io/my-workspace/

That URL is something you can actually send someone. It's not pretty - we'll fix that in Module 5 with Cloudflare - but it works and it's free.

Module 4 Local File vs Live Link: The 90-Second Mental Model

Before we get to Cloudflare, one concept you need to own. This trips up nearly everyone new to web work - and once you get it, it clicks forever.

GitHub gives you a free live URL (you just got one in Module 3). Cloudflare gives you a faster URL with custom domain support. Next module is where that happens.

Module 5 Cloudflare - The CDN That Replaces Hosting

You used GitHub to get your code online. Cloudflare is how you make it feel like a real website - custom domain, globally fast, SSL included, free. Most agencies pay $50-200/month for hosting setups that Cloudflare replaces for nothing.

What a CDN Actually Is

Set It Up - Three Commands

The Cool Stuff - What Agencies Actually Do With This

Here's where it gets useful for real client work. This is why Cloudflare matters beyond "free hosting".

Custom Domains - Brand It Like You Mean It

Instead of sending clients a URL like john-doe.github.io/project-name-thing-v2-final/, you send them graphicstudio.nyc. The mechanics: point your domain's DNS to Cloudflare (one setting, ~5 minutes), then in the dashboard connect that domain to your Pages project. First time takes about 15 minutes. Every time after, 5 minutes.

Subpages - One Domain, Many Client Projects

One domain can host multiple projects. Your agency at graphicstudio.nyc serves graphicstudio.nyc/client-alpha, graphicstudio.nyc/client-beta, graphicstudio.nyc/staging. Each path is a separate Pages project. You manage them all from one Cloudflare account. Clean, professional, $0.

Staging Links - Show Clients a Preview Before Going Live

Cloudflare Pages creates a unique preview URL for every git branch automatically. Your main branch is live on your custom domain. Your staging branch gets its own URL - something like staging-branch.project-abc.pages.dev. You send that to a client for approval. They click around, confirm it's right. You merge. The live site updates. Zero risk to what's already live.

Branded Client URLs - Stop Sending Ugly Links

What you send a client signals your professionalism. Compare:

• Ugly: https://yourname.github.io/client-project-v2-final-REAL/
• Professional: https://preview.graphicstudio.nyc/client-alpha

Two options: (1) set up a preview subdomain on your agency domain - costs nothing, (2) set up a domain on the client's own account and manage it for them - charges more, looks more premium.

Module 6Claude Code - Install, Auth, Permissions

Claude Code is the bridge between Claude's brain and your workspace. It reads your files, writes code, runs commands, and remembers context - all from inside VS Code.

Install the CLI

Verify it worked:

You should see a version number. If the command is not found, close and reopen VS Code.

VS Code Extension

If you installed the Claude Code extension in Module 2, you are already set. Open it with Cmd+L (Mac) or Ctrl+L (Windows). The extension automatically detects the CLI you just installed.

On first launch, Claude will ask you to authenticate. Click the link, sign in with your Claude account, and approve the connection.

Permission Modes

Claude needs permission to read files, edit files, and run commands. You control how much freedom it gets:

How to set it: Click the mode indicator at the bottom of the Claude panel. Or open VS Code settings (Cmd+,), search "Claude Code", and set your default mode.

Which Model to Use When

Claude comes in three sizes. Each has a sweet spot. Using the right model for the task saves money and gets better results:

Your First Week - Training Wheels

When You Get Stuck - How to Ask for Help

You will hit errors. Things will break. That is normal. Here is your troubleshooting playbook:

1. Screenshot the Error

On Mac: Cmd+Shift+4 to select the area. On Windows: Win+Shift+S. This saves a screenshot to your clipboard or desktop.

2. Paste It Into Claude

Claude Code can read images. Drag a screenshot directly into the Claude Code chat panel and ask: "What is going wrong here?" Claude will read the error message from the image and help you fix it.

3. Open Claude.ai as a Backup

If Claude Code is not responding or you are stuck in a way that needs a fresh perspective, open claude.ai in your browser. Same account, no extra cost. Paste the error text or upload the screenshot and ask for help.

4. How to Ask Effectively

When asking for help, include three things:

1. What you expected to happen - "I ran git push and expected it to upload my files."
2. What actually happened - "I got an error that says 'remote rejected'."
3. The error message - paste the exact text, or drag in a screenshot.

Module 7Your Identity Layer - CLAUDE.md + Memory

This is the module that separates "using Claude" from "having an AI business partner." Without this, Claude gives generic answers. With it, Claude knows your business, your role, your goals, and your preferences - every single session.

Claude.ai vs Claude Code - Which Is Which

Before setting up your identity layer, it helps to understand the two surfaces you will use every day and why each one needs separate context setup.

Training Claude.ai Memory

Claude.ai has a built-in memory system. Most people never set it up. Here's how.

• Where to find it: Settings → Memory, or create a Project and paste your context into the Project Instructions field
• What to put in: condensed version of your ChatGPT handover doc (300-500 words)
• Use the same answers as CLAUDE.md - just shorter
• Update it when your business situation changes significantly

CLAUDE.md - Your AI's Operating Manual

CLAUDE.md is a file at the root of your workspace that Claude reads at the start of every conversation. It tells Claude who you are and how to work with you.

Memory - Claude Remembers Across Sessions

Claude Code has a persistent memory system. Important facts about you, your preferences, and your projects are saved as small files that Claude reads at the start of each conversation.

How to save something to memory:

How to check what Claude remembers:

How to remove a memory:

Memory Hierarchy - 4 Layers

Claude Code has four layers of memory, from most permanent to most temporary:

Understanding the Context Window

Claude can "see" about 200,000 tokens at once - roughly 150,000 words. Think of it as Claude's short-term memory for the current conversation. Everything you say, everything Claude says, and every file Claude reads during the session takes up space in this window.

What Happens When the Window Fills Up

As your conversation grows, older messages get summarised or compressed to make room for new ones. This is automatic and usually seamless, but in very long sessions you may notice:

• Claude forgets instructions you gave earlier in the conversation
• Claude repeats itself or asks questions you already answered
• Claude loses track of the plan you agreed on 30 messages ago

How to Work With It, Not Against It

• Break long sessions into focused chats - one task per conversation works better than a 4-hour marathon
• Use memory files for persistence - anything Claude needs to remember across sessions should be saved to memory, not left in chat
• Use @file references instead of pasting - pointing Claude to a file uses less context than pasting its contents into the chat
• Start fresh when context gets stale - if Claude seems confused, open a new conversation. CLAUDE.md and memory carry over automatically.

Priming Your Memory on Day 1

Once your CLAUDE.md is written, run this prompt. It seeds your memory files with the same context so Claude has it in two places.

Rules + Skills (Optional, Powerful)

As you get comfortable, you can add two more layers:

• Rules (.claude/rules/) - domain-specific behaviour files. Example: "when writing client emails, always use this tone."
• Skills (.claude/skills/) - reusable multi-step workflows you trigger with /skill-name. Example: /monthly-report generates a full client report.

Your First Conversation - 5 Steps

You have Claude Code installed, authenticated, and your CLAUDE.md written. Here is your first real conversation:

Module 8API Connections - Top 10 + 5 Team Stacks

Without APIs, Claude can only read and write files in your workspace. With APIs, Claude becomes an operating system - reading your emails, updating your spreadsheets, posting to social media, pulling analytics, and managing your calendar. APIs are the difference between "smart notepad" and "AI-powered business."

The Top 10 - Start Here Regardless of What You Do

These 10 connections work for everyone. Regardless of your niche, business type, or how advanced you are - these are the foundation.

Below are 5 pre-built stacks. Pick the one closest to your work, then add individual tools from the others as needed. You do not need all five - scan the descriptions and choose the one that matches your daily workflow.

Stack 1: Social Media Team

For content creators, social media managers, and brand builders.

Stack 2: SEO Team

For SEO agencies, consultants, and anyone doing search engine work.

The SEO stack is data-heavy by nature. The payoff is that Claude can pull keyword data, cross-reference it with competitor backlinks, and draft a content brief - all in a single prompt.

Stack 3: All-in-1 Marketing Team

For agencies and marketers who need everything connected.

Stack 4: Personal Operator

For solo operators, freelancers, and anyone who wants Claude to manage their digital life.

The Personal Operator stack is the leanest - five connections and Claude handles your email, calendar, documents, and notes. Most solo operators find this is all they need for the first month.

Stack 5: Content Creator / Educator (6 APIs)

For course creators, YouTubers, educators, and anyone building an audience through content.

Security: .env Setup

API keys are passwords that let Claude talk to external services. They must never be committed to git.

Module 9Prompting That Actually Works + Ready-to-Use Prompts

Most people prompt like they are texting a friend. Vague. Incomplete. Then they wonder why the output is mediocre. Good prompting is about giving Claude the right context, constraints, and format - not about magic words.

The 5 Rules of Effective Prompting

Rule 1: Tell Claude What You Want, Not What You Don't Want

Rule 2: Give Context Before the Ask

Compare that to: "Write me a cold email." The first prompt gives Claude your niche, the prospect's situation, the specific gap, the format, and the tone. The output will be 10x better.

Rule 3: Specify the Output Format

Rule 4: Reference Files, Don't Re-explain

Claude Code can read files directly. Point it to the file instead of pasting content into the chat. This keeps context clean and uses your source of truth.

Rule 5: One Task Per Message (Usually)

Long multi-part messages create confusion. Break complex work into steps:

1. "Read the client profile and tell me what you understand."
2. "Now draft the report introduction."
3. "Add the recommendations section based on the gaps you found."

4 Prompt Patterns That Work

These four patterns cover the majority of useful prompting techniques. Learn them and you will get better results from any AI, not just Claude.

Pattern 1: Role Prompting

Tell Claude who to be. This focuses its knowledge and changes its tone.

Pattern 2: Chain of Thought

Force Claude to think before answering. This dramatically improves accuracy on complex questions.

Pattern 3: Few-Shot (Give Examples)

Show Claude what good looks like. This is the fastest way to get consistent output.

Pattern 4: Constraint Setting

Set boundaries so Claude does not wander. The more constraints, the more focused the output.

Data Extraction Patterns

Both Claude and ChatGPT are excellent at extracting, structuring, and transforming data. The key is being explicit about format:

Save Ideas on the Fly

Claude Code is your second brain. When ideas come to you mid-session, capture them instantly instead of losing them:

Claude Code Syntax

Claude Code has special syntax that regular Claude (browser) does not. Learn these three:

Now that you know how to reference files and write multi-line prompts, the next question is when to use Claude Code versus other tools you already know.

Claude Code vs ChatGPT - When to Use Which

Prompt 1 - The Business Quick Audit

Use this to analyse any business in 5 minutes. Works for prospects, competitors, or your own clients. Saves output directly to your workspace.

Prompt 2 - The Content System Starter

Use this to build a week of content in 10 minutes. Saves output directly to your workspace.

Your Daily AI Workflow (15 Minutes)

Build this habit and you will be ahead of 95% of people in your industry within a month:

Module 10You're Flying Now

You made it. Here is everything you now have:

Here is where to go next:

• Build your first skill - take a task you do weekly, write it as a skill file, and never do it manually again
• Add more API connections - each new MCP server is a new superpower
• Create rules for your domain - teach Claude your industry's conventions so every output is on-brand
• Learn agents - when you are ready, Claude can run multiple parallel tasks (like having a team of AI workers)
• Share your workspace - invite a business partner to your GitHub repo and both of you use Claude with the same context