This Claude Code guide for beginners walks through the four steps from installation to your first finished task: install the tool, start it in the terminal with claude, set up a CLAUDE.md as your first house rule, and let Claude work directly in your own files. You don't need to know how to code, you just need a terminal and a subscription.
I've run my business since 2026 on an AI workforce built on Claude, each with its own personnel file as a CLAUDE.md and its own skills. What started for me as a single install is now my daily terminal workspace. Everything I write about Claude is collected on the Claude page.
What Claude Code is, even if you don't code
According to Anthropic's official documentation, Claude Code is an "agentic coding tool that reads your codebase, edits files, runs commands." It's available in the terminal, in VS Code, in JetBrains IDEs, as a desktop app, and on the web at claude.ai/code.
The name suggests it's only for software development, but "code" here means any text file you want to work on: a proposal, meeting notes, a checklist, an entire folder structure for your business. What it does can be summed up in one sentence: you give it a task, Claude opens the relevant files, changes them, and reports back what it did.
Claude Code guide: requirements, cost, and installation
Claude Code needs a paid subscription, the free plan isn't enough according to Anthropic's pricing page. The current prices, as of September 2026, are summarized with sources in Claude Pricing and Plans Explained: Free 0 dollars (without Claude Code), Pro 20 dollars a month, Max 100 or 200 dollars a month, Team from 25 dollars per person. The desktop app that includes Claude Code also requires a paid subscription according to the docs.
Installation runs through a single terminal command according to Anthropic's documentation:
# macOS, Linux, or WSL
curl -fsSL https://claude.ai/install.sh | bash
# Windows PowerShell
irm https://claude.ai/install.ps1 | iex
The docs also list brew install --cask claude-code for macOS with Homebrew, or winget install Anthropic.ClaudeCode for Windows. After that, move into your working directory and start a session:
cd your-project
claude
That's the entire installation. What happens next is up to you, one prompt at a time.
Your first task: files instead of code
For your first task, I wouldn't start by restructuring an entire project. Open a folder with a few real files, notes or a proposal, for example, and ask for a summary of what's in there first. That's how you see Claude Code read before it writes.
In my own experience, Claude Code checks in before making larger changes, before it actually edits files or runs commands, that's my day-to-day experience using it, not a guarantee for every case. If you're unsure, ask it to show you a plan first before you approve a change. That caution early on pays off, especially when a task touches several files at once.
CLAUDE.md as your first house rule
Once your first task works, a CLAUDE.md is worth setting up. According to Anthropic's documentation, there are several levels: a managed policy, ~/.claude/CLAUDE.md for you as a user, ./CLAUDE.md or ./.claude/CLAUDE.md for a single project, and ./CLAUDE.local.md for local notes that aren't checked into version control. These files get merged, not overwritten, and the docs recommend keeping the project file under 200 lines.
Details on every level are in the Claude Code memory docs. Running /init has Claude Code generate a first CLAUDE.md for you, /memory opens the existing files for editing, and /context shows you what's currently loaded. Claude Code also keeps automatic notes under ~/.claude/projects/<project>/memory/. Important per the docs: CLAUDE.md is context Claude takes into account, not an enforced rule, hard restrictions belong in other mechanisms. I walk through a complete CLAUDE.md of mine in detail in CLAUDE.md: Claude Code's Memory.
Sebastian in the terminal: an SEO audit as an example
To show you what this feels like day to day, here's an example from my own workforce: Sebastian handles SEO and GEO for me, and he works exclusively as a Claude Code terminal employee, not through a web interface. The input is an assignment with a target URL, something like "check the technical SEO state of this website." Sebastian then sets up a working folder, calls tools like Lighthouse and his own crawling scripts from the terminal, collects the raw results as evidence, and writes a prioritized audit with findings and recommended actions from that.
What Sebastian doesn't decide himself: whether a report leaves the building or a fix actually gets deployed on a client site, I sign that off explicitly, and he records the approval with its timestamp in his log. The terminal employee researches and measures, the decision about anything client-facing stays with me. That split, research and execution in the terminal, approval by you, is for me the core of what Claude Code delivers for a business like mine.
Common beginner mistakes
The most common mistake I made early on was waving changes through without really reading the summary. Especially with several files at once, it's worth taking a moment to check what actually changed before you move on.
The second mistake is a missing or overly long CLAUDE.md: without one, you re-explain context every time, and with a file well past the 200 lines the docs recommend, it becomes unreadable itself. The third mistake is starting Claude Code in the wrong folder and then wondering why context is missing, claude reads your current working directory first, not your entire machine.
Frequently asked questions
Do I need to know how to code to use Claude Code?
No. Claude Code reads and edits any kind of text file, not just source code. For proposals, meeting notes, or folder structures in a consulting business, you don't need programming experience, just a terminal and some practice writing clear task descriptions.
What does getting started with Claude Code cost?
Claude Code isn't included in the free plan according to Anthropic's pricing page, you need at least a Pro subscription. The exact prices, as of September 2026, are listed with sources in Claude Pricing and Plans Explained, and I cover Claude Code's own usage costs in more depth in Claude Code Pricing: Plan or API?.
Can Claude Code delete my files?
Claude Code can change files and run commands, that's what it's built to do. In my own use, it checks in before larger changes, before anything is actually altered, but that's my personal experience, not a general guarantee. Especially early on, I'd read every summary and keep version control like Git on any folder that matters.
Does Claude Code run on Mac and Windows?
Yes, according to Anthropic's documentation, installation is available for macOS, Linux, WSL, and native Windows PowerShell, plus packages through Homebrew and winget. The exact commands are in the installation section above.
What's the difference from Claude Cowork?
Both build on the same foundation, things like CLAUDE.md and skills, but they differ in interface and target audience. I've laid out the full comparison in Claude Code or Cowork: What For?, a short answer here wouldn't do the question justice.
Where you go from here
After installing and setting up your first CLAUDE.md, the next sensible step is turning your recurring tasks into Claude Skills: How to Create and Use Them, so you don't have to describe them from scratch every time. If you're unsure whether you need Claude Code or the simpler Cowork interface, the comparison in Claude Code or Cowork: What For? settles that.
The eight ready-made AI employee packages, each with its own personnel file and its own skills, are in my community. My suggestion for getting started: install Claude Code today, open a small, real folder, and ask for a summary of what's in it first, before you approve your first change.
