How to Install Claude Code in 2026: Zero to "Hi, I'm Claude" in 20 Minutes (No Code Editor)

By Jonah, Founder of FindClout — July 2026

I run a distribution network that has generated over 3.3 billion views. I am not a software engineer. I never opened a computer science textbook. And I build most of the internal tooling that runs FindClout myself, in a terminal, by typing English at an AI agent named Claude and hitting enter.

This guide is the exact setup I hand to every head of growth and non-technical founder who asks me "how are you shipping tools this fast?" The honest answer is: I installed one program, and the program does the coding. The only skill you need to bring is the willingness to paste an error message into a chat window and try again. That's it. That is the whole job.

By the end of this you will have Claude Code running on your machine, a two-letter shortcut to launch it, and a first thing to build. Budget 20 minutes. Budget 40 if the gods of your operating system decide to be difficult — which is normal, and which we will handle without a single Google search.

What Claude Code actually is

Claude Code is Anthropic's command-line agent. "Command-line" means it lives in your terminal — that black window with text that you have been quietly afraid of your entire career — instead of a browser tab. "Agent" means it doesn't just talk. It does. It creates files, edits them, runs commands, installs software, reads the errors it just caused, and fixes them, in a loop, until the thing you asked for exists.

That last part is the whole game. A chatbot in a browser gives you a block of code and says "paste this into your project." Claude Code is the thing pasting it in, running it, watching it break, and repairing it — while you drink your Celsius and watch. It is the difference between an assistant who emails you instructions and an assistant who sits down at your keyboard and finishes the task.

Why terminal beats the browser chatbots for real work

If you've been living in ChatGPT, Claude.ai, Lovable, Replit, or v0, you already know the ceiling: they're great for a snippet or a landing page, and then they hit a wall. The moment your project has more than a couple of files, or needs to talk to a real API, or has to run on a real server, the "copy the code out of the chat by hand" workflow collapses under its own weight.

The terminal removes the copy-paste middleman. Claude Code has direct access to your files and your machine, so it can build a project with fifty files as easily as one, wire up real API keys, and actually run what it wrote to confirm it works. This is why builders who graduate beyond Lovable and Replit almost all end up in the terminal. It's not that the toy tools are bad — it's that you outgrow them the second you want real capability. If you want the fuller picture of that jump, read what a growth engineer is — it's the role this setup turns you into.

Prerequisites: what you need before you start

Almost nothing, and the little you need, Claude will help you install. Here's the honest list.

• A computer. Mac, Windows, or Linux all work. I'll call out the Mac and Windows specifics as we go.
• A Claude subscription. This is what powers the agent. I recommend the Claude Max plan — more on why below. You'll sign in during setup; you do not need to touch an API key to get started.
• Node.js. Claude Code is distributed as an npm package, and npm comes with Node. If you don't have Node, don't panic — installing it is step one, and we do it together.
• claude.ai open in a browser tab. This is your install copilot and the single most important tool in this entire guide. Do not skip it.
• A can of Celsius. Optional, but the guide it comes from insists, and I've learned not to argue.

Notice what's not on that list: a code editor (VS Code, none of it), a GitHub account, any prior programming knowledge, or the ability to read the code that gets written. You need none of those things to start. That is not me being cute — that is the actual bar.

Step 1: Make a home for your work

Before anything technical, do the least technical thing imaginable. On your desktop, make a new folder and call it projects. This is where everything you build will live. Keeping it in one dedicated folder isn't just tidiness — it's a safety boundary that matters a lot once you start letting the agent run freely, which I cover in the companion piece on what --dangerously-skip-permissions actually does.

Now the small piece of magic that makes everything downstream painless: right-click the folder and choose the option to open a terminal at that folder.

• On Mac: right-click the folder → Services → New Terminal at Folder. (If you don't see it, it's a one-time toggle in System Settings → Keyboard → Keyboard Shortcuts → Services — and yes, you can ask claude.ai exactly where the toggle is.)
• On Windows: right-click inside the folder → Open in Terminal.

A terminal window opens, already sitting inside your projects folder. You didn't have to type a single navigation command. This right-click-to-terminal move is the one habit that makes the whole no-code-editor lifestyle actually work, so burn it into your muscle memory now.

Step 2: Meet your install copilot (this is the real trick)

Here is the mindset shift that makes this a 20-minute job instead of a two-day rage spiral. You are not going to install Claude Code by following steps perfectly. You are going to install it by failing, pasting the failure into a chat, and following what it says — over and over — until it works.

Open claude.ai in your browser. This tab is now your install copilot. The rule is dead simple, and I want you to tattoo it somewhere:

The Brute-Force Rule: Any time the terminal spits out an error — any error, no matter how scary or how long — select the entire thing, copy it, paste it into claude.ai, and add "I'm installing Claude Code, what do I do?" Then do exactly what it tells you. If that produces a new error, repeat. You cannot lose this game. You can only quit.

This works because installation errors are the most common, most-documented problems in all of software. There is no error you will hit that the model hasn't seen ten thousand times. Your job is not to understand the error. Your job is to be the hands that copy it over and paste the fix back. Stop trying to comprehend, start trying to complete.

Step 3: Get Node.js (npm) on your machine

Claude Code installs through npm, Node's package manager. Let's check whether you already have it. In your terminal, type this and press enter:

node --version

Two things can happen. If you see something like v20.11.0, congratulations, you have Node and you can skip to Step 4. If instead you see something like this:

zsh: command not found: node

Then Node isn't installed yet. You know the drill: that error goes straight into claude.ai. But here's the fast path so you know what "good" looks like.

On Mac: use Homebrew

The cleanest way to get Node on a Mac is a tool called Homebrew. Paste the following into claude.ai — "walk me through installing Homebrew and then Node on my Mac" — and it'll hand you the exact commands to run. They look roughly like this:

# Install Homebrew (the copilot gives you the current one-liner)
/bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/HEAD/install.sh)"

# Then install Node
brew install node

Homebrew's installer sometimes ends by telling you to run two more lines to "add Homebrew to your PATH." If it does, that's not an error — it's an instruction. Copy those exact lines, run them, done. If any of it goes sideways, you already know: copy the red text, paste it into claude.ai, follow along.

On Windows: grab the installer or use WSL

On Windows the simplest route is downloading the Node.js installer from the official site and clicking through it. The more powerful route — the one I'd steer you toward if you plan to build seriously — is installing WSL (Windows Subsystem for Linux), which gives you a proper Unix-style terminal that behaves like a Mac's. Ask claude.ai "should I use WSL for Claude Code on Windows and how do I set it up?" and let it guide you. Either way works; WSL just has fewer sharp edges long-term.

When node --version finally prints a version number, you've cleared the hardest gate. Everything after this is easier.

Step 4: Install Claude Code

This is the moment you came for, and it's anticlimactically short. One command:

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

The -g means "install globally" — so Claude Code is available from any folder, not just this one. Press enter and let it run. You'll see a stream of text scroll by as npm downloads and wires things up. That scrolling text is normal; it is not errors.

Two common bumps here, both trivially solved:

• A wall of permission errors mentioning EACCES or sudo. This means npm doesn't have permission to write to the global folder. Do not just blindly slap sudo in front of it — paste the error into claude.ai and let it give you the clean fix (usually a one-time npm config change). This is exactly the kind of thing the copilot exists for.
• It finishes with no obvious errors. Great. It worked. Move on.

Verify it landed:

claude --version

A version number means Claude Code is on your machine. If it says "command not found," that's a PATH hiccup — you guessed it, into the copilot it goes, and you'll be sorted in two minutes.

Step 5: The cc alias — the two letters that change everything

Typing claude every time is fine. But there's a shortcut that the people who build fastest all use, and it does two jobs at once: it shortens the command to two letters, and it turns on the flag that lets the agent actually flow. Here's the alias:

alias cc="claude --dangerously-skip-permissions"

That --dangerously-skip-permissions flag stops Claude Code from stopping to ask you "may I run this command?" every thirty seconds. For an engineer reviewing every line, the prompts make sense. For a non-technical builder who just wants the tool built, being interrupted constantly kills the whole point. I wrote an entire honest breakdown of the trade-off — what it turns off, why velocity-focused builders turn it on, and the real risks — in this companion post. Read it before you go wild with it; the short version is "use a dedicated projects folder and don't point it at anything you can't afford to lose."

To make the alias permanent so you don't retype it every session, it needs to go into your shell's config file (~/.zshrc on a modern Mac, ~/.bashrc on many Linux/WSL setups). Don't memorize that. Paste this into claude.ai: "add alias cc='claude --dangerously-skip-permissions' to my shell config permanently, I'm on a Mac." It'll give you the one line to run. Then either open a new terminal or run source ~/.zshrc to load it.

Step 6: First launch — get it to say hi

Everything's installed. Time to wake it up. In your terminal, type your new alias and hit enter:

cc

The first time, Claude Code walks you through signing in. It kicks off a /login flow that opens your browser, you authorize with your Claude account, and you bounce back to the terminal. This is where your Claude Max subscription plugs in — your usage runs against your plan instead of a per-token meter, which for daily building is exactly what you want.

Once you're signed in, the agent is live and waiting. Type a greeting:

> hey, are you there?

And it answers — introduces itself, tells you it's Claude, and asks what you'd like to build. That's the "Hi, I'm Claude" moment. The terminal you were afraid of is now the most capable employee you've ever had, and it starts at whatever your subscription costs, working at 3am without complaint.

What to build first

Do not try to build the next unicorn on day one. The point of your first build is to feel the loop — describe, watch it work, see a real result — so you trust the tool. Good first projects for a growth person:

• A single-page tool for a repetitive task you do by hand. "Build me a webpage where I paste a list of URLs and it de-dupes them and sorts them." Boring, useful, done in two minutes.
• A tiny dashboard that reads a CSV. "Make an HTML page that loads this spreadsheet and shows me the top 20 rows by view count as a bar chart." Now you have a report generator.
• A script that renames or reorganizes files. Painful by hand, trivial for the agent.

Notice the pattern: each of these takes something you currently do manually and hands it to the machine. That's the entire premise of becoming a growth engineer — you stop waiting on an eng team's roadmap for internal tools and just build them. Once you're comfortable, the next moves are wiring up real capability: a Cloudflare account so the agent can deploy things to the internet, OpenRouter to route grunt work to the cheapest models, and social/ads data APIs so it can pull real numbers. All of that is laid out, step by step, in the guide.

The whole thing in one breath

1. Make a projects folder, right-click → open terminal at it.
2. Open claude.ai in a browser — this is your install copilot.
3. Get Node.js (node --version to check; Homebrew on Mac, installer or WSL on Windows).
4. Run npm install -g @anthropic-ai/claude-code.
5. Set alias cc="claude --dangerously-skip-permissions" and make it permanent.
6. Type cc, sign in with your Claude Max plan, and say hi.
7. Any error at any step → copy it → paste into claude.ai → do what it says → repeat.

That's it. That's the on-ramp that turns a marketer into someone who ships software. I've watched people who "aren't technical" go from this install to running their own internal tooling inside a week. The install feels like the hard part. It's actually the last easy part — after this, you're just describing what you want and watching it appear.

Jonah is the founder of FindClout, a curated creator distribution network that has generated 3.3B+ views for brands across sports, prediction markets, AI, and more. He builds most of FindClout's internal tooling himself, in a terminal, with zero formal coding background. Reach him at [email protected] or book a call.