Most guides for installing Claude Code assume you already know what you're doing. They throw commands at you without explaining why, skip over the parts that actually trip people up, and leave you staring at a terminal wondering if you've broken something.
This guide is different. Whether you're a business owner evaluating Claude Code for your operations, or someone technical who just wants the friction points documented honestly, this walks through the full setup. What it is, what you need, every step of the installation, and what to do when things go sideways.
By the end you'll have Claude Code running and a clear sense of whether it's the right tool for what you're trying to do.
What You're Actually Installing (And Why It Lives in Your Terminal)
Claude Code is Anthropic's agentic coding tool. It runs in your terminal, the command-line interface on your computer, rather than in a browser window. That's not just a technical detail. It's the whole point.
Because it lives in your terminal, Claude Code can read and write files on your machine, run commands, interact with your codebase, and take actions across your actual system. It's not a chat window. It's closer to a developer who sits alongside you and can touch the real work.
Claude Code vs the Claude chat interface — what's different
Claude.ai is a conversation. You type, it responds, nothing changes on your computer. It's great for drafts, explanations, brainstorming, summarising documents you paste in.
Claude Code is an agent. It can open a file, read what's in it, make changes, run the result, check if it worked, and loop back if it didn't. It operates on your actual project, not on text you've copied into a chat box.
The browser version has no access to your file system. Claude Code does. That's the fundamental difference.
What Claude Code can do that the browser version can't
A few things that only work in Claude Code:
Read and edit files in your project directory
Run terminal commands and see the output
Navigate a codebase, understanding how files relate to each other, not just reading one file at a time
Carry out multi-step tasks autonomously: write the code, test it, fix the error, test again
Interact with tools like Git, your package manager, or any CLI tool you have installed
At AMPL, we use Claude Code as the build environment for client automation systems. When we're building a custom AI workflow for a business, Claude Code handles the actual implementation work. It reads the existing setup, writes the integration code, runs tests. It's not a helper. It does the work.
Before You Start: What You Need
Three things need to be in place before you install Claude Code. Getting these right upfront saves a lot of headache later.
System requirements (Windows, macOS, Linux)
Claude Code runs on macOS, Linux, and Windows via WSL (Windows Subsystem for Linux). If you're on a Mac or Linux machine, you're already in the right environment. If you're on Windows, you'll need WSL set up first. Microsoft's own documentation covers WSL installation. It takes about ten minutes and is a one-time setup.
There's no minimum spec that's unusual. If your machine runs modern software comfortably, it'll run Claude Code.
Anthropic account and API access
Claude Code authenticates through your Anthropic account. You'll need an account at console.anthropic.com and you'll need API access enabled. This is not the same as a Claude.ai subscription. The API is a separate thing billed on usage.
Go to console.anthropic.com, create an account if you don't have one, and grab your API key from the settings. Keep that somewhere handy. You'll need it during setup. Don't share it or commit it to a repository.
Anthropic offers some free credits when you sign up. For evaluation purposes, you're unlikely to burn through them quickly.
Node.js — why it's needed and how to get it
Claude Code is installed via npm, which is the package manager that comes with Node.js. If you don't have Node.js installed, npm won't exist on your machine and the installation command won't work.
Go to nodejs.org and download the LTS version (Long Term Support). That's the stable one. Not the latest, the LTS. Install it, restart your terminal, and run node --version to confirm it's there. You should see something like v20.x.x or higher.
Node.js is just infrastructure here. You're not writing Node applications. It's just the system Claude Code needs to run.
Step-by-Step Installation
Once the prerequisites are in place, the installation itself isn't complicated. Here's each step with context on what's actually happening.
Step 1 — Install via npm
Open your terminal and run:
The -g flag installs Claude Code globally, meaning you can run it from any directory on your machine, not just the folder you're currently in. That matters because you'll want to use it across different projects.
The installation downloads the package and its dependencies. Depending on your connection, this takes a minute or two. You'll see a lot of text scroll past. That's normal.
Step 2 — Authenticate with your Anthropic account
Once installed, run:
The first time you run this, Claude Code will ask you to authenticate. It'll open a browser window, or give you a URL to open, pointing to Anthropic's authentication flow. Log in with your Anthropic account credentials.
Alternatively, if you want to authenticate using your API key directly, you can set it as an environment variable:
For most people evaluating Claude Code, the browser authentication flow is simpler. The environment variable approach is better for production systems or CI/CD pipelines where interactive login isn't an option.
Step 3 — Verify the installation
Run this to confirm everything is working:
You should see a version number. If you see an error, something went wrong with the installation. Most likely a Node version issue or a permissions problem. Both are covered in the troubleshooting section below.
Step 4 — Run your first command
Navigate to a folder you want to work in, your project directory or even just a test folder:
This opens the Claude Code interactive session. You're now in. You can type a request in plain English: "What files are in this directory?" or "Read the README and tell me what this project does." Claude Code will respond and, if appropriate, take action.
You're not writing code at this point. You're talking to a system that can then write and run code on your behalf.
Common Installation Problems and Fixes
These are the three issues we see most often. The official docs don't cover these in much detail, which is frustrating when you're stuck.
Permission errors on macOS
If you see something like EACCES: permission denied when running the npm install command, it usually means npm is trying to write to a directory it doesn't have access to.
The clean fix is to use a Node version manager like nvm instead of the system Node. Install nvm, use it to install Node, then run the Claude Code installation again. This sidesteps the permissions issue entirely because nvm installs Node in your home directory where you have full access.
Avoid using sudo npm install -g as a fix. It works but it creates permission problems down the line with other packages.
Node version conflicts
Claude Code requires a reasonably recent version of Node. If you installed Node a few years ago and never updated it, you might run into compatibility issues.
Run node --version. If you're below Node 18, update. If you're using nvm, it's just nvm install --lts and nvm use --lts. If you installed Node directly, download the current LTS from nodejs.org and reinstall.
This is one of those friction points that's obvious in hindsight but catches people out, especially on machines that have been around for a few years.
Authentication not persisting
Some people find that after authenticating, Claude Code asks them to log in again next session. This usually happens when the credentials aren't being stored properly. Often on systems with unusual home directory configurations or on shared machines.
The fix is to set your API key as a persistent environment variable. On macOS/Linux, add this to your shell profile file (~/.zshrc or ~/.bashrc):
Then run source ~/.zshrc, or restart your terminal, and you won't need to authenticate again.
Your First Real Task: A Practical Walkthrough
Once Claude Code is running, here's a practical first task that shows you what it can actually do, without needing an existing codebase.
Create an empty folder, navigate to it in your terminal, and start Claude Code:
Now ask it something concrete. For example: "Create a Python script that reads a CSV file and outputs a summary of the columns — column names, number of rows, and any missing values."
Watch what happens. Claude Code will write the script, create the file, and quite possibly ask if you want it to create a sample CSV to test with. If you say yes, it'll do that too, then run the script against the sample data.
You haven't written a line of code. You haven't opened a separate editor. The whole thing happened in the terminal, driven by plain English.
That's the point. The value isn't "Claude Code is clever." The value is that a non-developer can direct it toward a real task and get a working result. And a developer can hand it repetitive implementation work and focus on the harder problems.
For business owners evaluating Claude Code as part of a broader AI system, this first task gives you a clear sense of how the tool operates. Worth doing before making any decisions about how you'd integrate it into your operations.
What to Do Next After Setup
Getting Claude Code installed is the easy part, to be honest. Getting it working well for your specific situation takes a bit more thought.
A few things worth exploring once you're set up:
CLAUDE.md files. You can create a file called
CLAUDE.mdin any project directory and Claude Code will read it automatically at the start of each session. Use it to give context about the project. What it does, conventions to follow, things to avoid. This is how you get consistent behaviour across sessions.Slash commands. Type
/helpin a session to see what's available. There are shortcuts for common actions that speed things up once you're using Claude Code regularly.Project-level vs global configuration. Claude Code has settings you can configure both globally, applying to all projects, and locally, applying to a specific project. Worth understanding the difference once you're past the basics.
If you're evaluating Claude Code for a business context, thinking about how it could fit into your operations rather than just experimenting personally, the question worth asking is: what are the repetitive, structured tasks in your business that someone currently does manually? Claude Code doesn't replace your team. It removes the low-value work that slows them down.
If that question is hard to answer without more information, that's actually what an AMPL audit is for. We map your operations, identify the highest-value automation opportunities, and give you a clear picture of what's worth building before you commit to anything. If that's useful, book a free audit at amplconsulting.ai.
FAQ
Do I need to know how to code to use Claude Code?
Not to use it, no. You give instructions in plain English and Claude Code handles the implementation. That said, having some technical literacy helps. Knowing roughly what a file system is, how a terminal works, what an API key is. You don't need to write code. You do need to understand enough to point the tool in the right direction.
How much does Claude Code cost?
Claude Code is billed based on API usage, the tokens processed per session. Anthropic provides free credits when you sign up, which covers experimentation. For regular use, costs vary depending on how intensively you use it. For context, light usage by a single person is unlikely to cost more than a few dollars a month. Heavy usage in production systems is a different calculation.
Is Claude Code safe to use with my actual business files?
Claude Code only has access to directories you point it at. It doesn't crawl your machine autonomously. Treat it like you'd treat any tool with file access. Don't run it in directories containing sensitive credentials or data you wouldn't want touched. Use a dedicated project folder. Be specific in your instructions about what you want it to do.
What's the difference between Claude Code and the Claude API?
The Claude API is the underlying interface for building applications that use Claude programmatically. Claude Code is a pre-built tool that runs in your terminal and uses that API under the hood. Most people evaluating Claude Code for business use don't need to interact with the API directly. Claude Code is the layer that makes the API accessible without custom development.
Can I use Claude Code on a team?
Each person on a team installs Claude Code individually with their own authentication. There's no shared session or real-time collaboration built in. For team use in a business context, the more interesting question is usually about shared project configuration. Using CLAUDE.md files to give everyone consistent context, and how Claude Code fits into your broader workflow. If you want setup guidance specific to your situation, that's worth a conversation.
What if my installation still isn't working?
Check three things in order: Node version (run node --version, needs to be 18 or above), permissions (use nvm if you're getting EACCES errors), and authentication (set your API key as an environment variable if credentials aren't persisting). If none of those resolve it, Anthropic's Discord community is active and the support team is responsive.
