In short. Claude Code is Anthropic’s assistant that works directly on a project’s files: it reads a folder, proposes a plan, edits the files and runs commands, asking for confirmation before important actions. You pay for it with a Claude subscription or through the API. It makes sense even for people who don’t code, if they have a concrete project and the time to check the work.
What is Claude Code, in plain words?
It’s Claude, Anthropic’s model, put to work inside a folder instead of inside a chat. You tell it in plain language what you want to achieve. It reads the files that are there, tells you how it plans to proceed, changes what’s needed and shows you the before and after. If a command needs running, such as publishing a site or running a check, it does that too.
The chat gives you an answer to copy. Claude Code does the work, and you review it.
Where do you use it?
| Where | How it works | When I use it |
|---|---|---|
| Terminal | You open it in a folder on your computer and write to it | When I work on files that live only on my PC |
| Desktop app | The same thing, with a windowed interface | For people who don’t want to use the terminal |
| Web and mobile app | It works in a cloud environment connected to a GitHub repository | For websites: every approved change goes live by itself |
| Scheduled routines | An instruction that starts at a fixed time | Every morning, for my sites’ content calendar |
How much does Claude Code cost?
There are two ways to pay for it:
- Claude subscription. Claude Code is included in Claude’s paid plans, with usage limits that grow with the plan.
- Anthropic API. You pay per use, based on how much text Claude reads and writes. It makes sense for companies or automated uses.
I don’t give figures here because plans and prices change often: the current ones are at https://claude.com/pricing. The individual plan is enough to start; if you use it every day across several projects, the limits become the thing to watch.
Who does it make sense for if you’re not a developer?
| Profile | Does it make sense? | Why |
|---|---|---|
| Product managers | Yes | Real prototypes instead of mockups, and a concrete way to understand what to ask the team for |
| People with a website or a small static e-commerce site | Yes | Copy, new pages and SEO checks without waiting for someone |
| People who manage data in spreadsheets and files | Yes, with care | Scripts that tidy up or check repetitive files; the data must always be verified |
| People who want a complex app to sell | Only as a start | A prototype, yes; security, payments and maintenance need a developer |
| People who only want answers or copy | No | Claude’s chat is enough |
What have I built with it?
I don’t write code by hand, yet today with Claude Code I run:
- This site. HTML pages on GitHub, published automatically to Firebase with every change. The full story is in How I built my site with Claude Code.
- The sites for my rentals and the one for Staged.ai, with guides and articles that come out on a schedule.
- A control page that brings together the sites’ Search Console, Analytics and Cloudflare data and decides what to publish. Every morning a Claude Code routine reads it and does the day’s work.
The part that matters isn’t the code. It’s well-written instructions: what it can do, what it mustn’t touch, how to check that the result is right.
What doesn’t it do, or do badly?
- It doesn’t know your data unless you give it. If a number isn’t in the files, it may write a plausible one. I always check the numbers myself.
- It doesn’t replace judgement. It does what you ask, even when it’s a bad idea. Asking it for a plan first helps you stop in time.
- It isn’t magic on large projects. The bigger the project and the fewer the rules, the more time you spend correcting.
- Keep it away from secrets. Passwords and keys don’t belong in the folders you let it read or in repositories.
How do I start, in practice?
- Pick a small, real project: a page, a file to tidy up, a check you do by hand every week.
- Create a folder and write a short CLAUDE.md file: who the audience is, what tone to use, what not to touch.
- Ask for the plan first, then the changes, and look at what changed before you approve.
- When it works, connect it to GitHub: every change becomes traceable and reversible.
If you want to bring this way of working into a team, I write about it in The method and in Learning AI at work.
Frequently asked questions
Is Claude Code free?
No. You use it with a paid Claude subscription or with the Anthropic API, paying per use. Plans and prices change often: the current ones are at https://claude.com/pricing.
What’s the difference between Claude Code and Claude’s chat?
The chat answers and gives you text or code to copy. Claude Code works inside a folder: it reads the files, edits them, runs commands and shows you what it changed. It’s the difference between advice and a colleague who does the work in front of you.
Do you need to know how to code to use Claude Code?
For small projects, such as a static site or a script for your own files, no. You do need to describe clearly what you want and check the result: Claude Code asks for confirmation before the actions that matter, and the decision stays yours.
Does Claude Code work on Windows?
Yes, as well as on macOS and Linux. It’s also available in the desktop app and on the web, where it runs in a cloud environment connected to GitHub.
The examples contain only data from my own projects. No data from clients, guests or real accounts. Published on 12 November 2026.