How to install Claude Code on Mac and Windows — no developer experience required.

What is Claude Code and who is it for?

Claude Code is Anthropic's command-line AI coding agent. Unlike Claude in the browser, Claude Code runs on your own computer — it can read files, edit code, run terminal commands, and carry out multi-step tasks without you having to copy-paste anything between windows. It's built for developers, but the install process is the same for anyone, and you don't need to write code to get value out of it immediately.

The most common non-developer uses: editing and improving existing HTML or document files, running optimization scripts against your own content, automating repetitive file tasks, and directing it to make changes to a website or project one plain-English instruction at a time. The agent handles the technical steps; you provide the goal.

What separates Claude Code from the web interface is that it has direct access to your filesystem. It can open a file, rewrite it, save it, and run a check — all in one instruction. That loop — read, change, verify — is what makes it genuinely useful for optimization work rather than just answering questions.

How do I install Claude Code on Mac?

Open the Terminal app and run one command: curl -fsSL https://claude.ai/install.sh | bash. That's the native installer Anthropic now recommends, and it needs no Node.js and no Homebrew. The whole thing takes about 5 minutes the first time. Per Anthropic's official setup docs (verified June 15, 2026), Claude Code requires macOS 13.0 or later and 4 GB+ of RAM.

Step 1 — Run the installer

Open the Terminal app (find it in Applications → Utilities, or press Command + Space and type "Terminal"). Paste this command and press Enter:

The installer downloads the Claude Code binary, puts it on your system path, and keeps itself updated in the background. Prefer Homebrew? brew install --cask claude-code works too (you'll update it yourself with brew upgrade claude-code).

Step 2 — Verify and launch

Confirm it installed by typing claude --version and pressing Enter — you should see a version number. Then type claude and press Enter to start it. The first time, it opens a browser window to log in with your Anthropic account (or you can paste an API key). If anything looks off, claude doctor checks your setup.

Alternative: install with npm

If you already have Node.js (version 18 or later, per Anthropic's docs) or prefer the package-manager route, you can install Claude Code globally with npm instead:

Don't have Node? Install it first with Homebrew (brew install node), then run the npm command above. Either way, you finish by typing claude to launch. Anthropic advises against sudo npm install -g — if you hit permission errors, the native installer above sidesteps them entirely.

How do I install Claude Code on Windows?

Open PowerShell and run irm https://claude.ai/install.ps1 | iex. Claude Code now runs natively on Windows — you no longer need WSL. Per Anthropic's setup docs (verified June 15, 2026), native Windows requires Windows 10 (build 1809) or later; WSL is still supported and is the path to choose if you want a Linux toolchain or sandboxed command execution.

Option A — Native Windows (recommended)

Open PowerShell (search for it in the Start menu — you do not need to run it as Administrator) and run:

That downloads and installs Claude Code, then keeps it updated automatically. When it finishes, type claude and press Enter to launch, and follow the browser login prompts. Installing Git for Windows is optional but recommended — it lets Claude Code use the Bash tool; without it, Claude Code uses PowerShell for shell commands.

Option B — WSL (for Linux toolchains)

If you want a full Linux environment inside Windows, open PowerShell as Administrator and run wsl --install, then restart when prompted. Open the Ubuntu app that appears in your Start menu, complete its setup, and inside the Ubuntu terminal run the same native installer as Mac and Linux:

Then type claude to launch and follow the login prompts. You run and launch Claude Code inside the WSL terminal, not from PowerShell.

What happens on your first run, and how do you enable auto mode?

You land in an interactive session scoped to your current folder, and you turn on auto mode by approving permissions when Claude Code asks. The first time you type claude and log in, Claude Code reads your current directory — the folder your terminal is in — as the working context. Everything it does is relative to that folder unless you tell it otherwise.

By default, Claude Code runs in a cautious mode where it asks permission before taking actions (reading files, making edits, running commands). This is sensible for first use, but for real optimization work you want auto mode, where it can carry out a full instruction end-to-end without interrupting you at every step.

The recommended way to enable auto mode is to grant permissions through the in-session prompt when Claude Code asks — you can allow a tool for the rest of the session in one keystroke. There is also a blanket --dangerously-skip-permissions flag that skips every confirmation; the name is Anthropic's own warning, so reserve it for a throwaway copy of a project you can afford to lose. For everyday editing of your own local HTML and content files the risk is low — Claude Code is changing files just as a text editor would, and you can undo any change with git. For anything touching live servers, production databases, or external systems, review each proposed action first.

A good first prompt to test that everything is working: navigate to a folder that has a text or HTML file in it, type claude, and then at the prompt write something like: "Read the file landing-page.html and list the five things you'd improve about the headline section, in order of expected impact." You should see Claude Code read the file and give you a specific, file-grounded response — not a generic answer. If it does, your install is working correctly.

Useful Claude Code commands to know

What should marketers build first with Claude Code?

Start by pointing it at one real file with a concrete goal — a landing page to improve, an email sequence to audit, or a keep-the-winner optimization loop. The most common mistake after installing Claude Code is using it like a chat assistant — asking it questions and reading the answers. That misses its advantage entirely. Claude Code is most powerful when you give it a goal and a file, and let it take the action itself.

Here are the three starting projects that produce immediate results for non-developers:

• Landing page headline optimizer. Navigate to your site's HTML file and ask Claude Code to generate five headline variants for a specific section, explain the hypothesis behind each, and output them as a ready-to-test list. You evaluate, pick the best two, tell it to swap them in.
• Email sequence auditor. Put your welcome emails in a folder and ask Claude Code to read all of them, identify the weakest CTA in each, and suggest a replacement with a brief rationale. No code required — it's working on plain text files.
• Autoresearch loop runner. This is the most powerful use: a program.md instruction file that tells Claude Code to propose a change, apply it, score it against a baseline, keep it if it wins and revert if it doesn't — then repeat. The Playbook has 12 templates with this loop pre-written.

The common thread: you write a plain-English instruction, Claude Code handles the implementation. Your job is to define the goal and evaluate the outputs — the agent does the iteration. That division of labor is the whole point, and the Autoresearch Playbook gives you 12 loop templates that already encode the right instruction structure, scoring logic, and safety limits so you're not starting from a blank program.md.