Claude Code is a tool made by Anthropic, the company behind Claude, that builds and changes software for you. You type what you want in plain language. It reads the files in a folder on your computer, edits them, runs the commands a developer would run, checks the result, and reports back. It is the same Claude you may have chatted with, given the ability to act instead of only answer. It is not the chat app, it is not a website builder like Lovable or Base44, and it does not host anything. It comes with a paid Claude subscription, and there is no free tier of Claude Code itself.
Claude and Claude Code are two different things
“Claude” is the AI, and also the app you talk to it in: the website, the phone app, the desktop app. You ask, it answers. If you ask it for an app, it hands you code in a box and leaves the rest to you: making the files, putting them somewhere, running them. That is where most people who cannot code get stuck.
Claude Code is a separate tool that uses the same AI but works inside your project. Anthropic’s own description is a tool that “reads your codebase, edits files, runs commands, and integrates with your development tools.” In practice the files appear in your folder and the commands run without you. Your job is to say what you want and to check what came back.
The two share one subscription and one budget. Anthropic counts everything you do in the chat app, the desktop app and Claude Code against the same usage limit. The Claude desktop app makes the split visible: it has a Chat tab, a Cowork tab for background tasks, and a Code tab, and the Code tab is Claude Code. The free Claude plan gives you the chat. Claude Code needs a paid plan.
What happens when you ask it for something
Anthropic describes the work as a loop with three parts: gather context, take action, verify results. When you ask for a change, it first reads whatever it needs to understand your project. Then it acts: creates or edits files, runs a command, starts the app. Then it checks, by running the app or a test, and reads what came back. You can interrupt at any point to steer it.
How much it does without asking depends on a setting called the permission mode. In Manual mode it asks before every file change and every command. On most paid plans the starting mode is called Auto, where a separate checker reviews each action in the background and blocks the risky ones instead of asking you. There is also a Plan mode, where it reads and proposes but changes nothing, which is the one for a big change.
Two safety nets are built in. Before it edits a file it saves a copy, so a bad change can be rewound. And it works with Git, the standard tool for keeping versions of a project, so a working version is never lost. Each new session starts with a fresh memory, so it reads a plain text file in your project, called CLAUDE.md, at the start of every session for the facts about your app.
What Claude Code is not
It is not a website builder. There is no preview pane, no Publish button, no built-in database, no built-in login. Lovable gives you those in the browser. Claude Code gives you files, and you add the rest as you need it, one piece at a time.
It is not a hosting service. Your app runs on your computer while you build it, and putting it on the internet is a separate step with a separate service. Claude Code does the work of that step with you, but it does not provide the place.
It is not paid for by credits. Lovable and Base44 sell credits per message. Claude Code is metered by time: you get an allowance that refills over a set window, shared with the chat app. When you hit the limit you wait for the reset, or on a paid plan you can switch on usage credits, which are billed separately and on top of the subscription. One trap: if an Anthropic API key is set on your computer, Claude Code bills that key instead of your subscription, at usage rates. Anthropic’s help pages say so in plain words.
It is not one program. It runs in a terminal, the text-only window every computer has, inside the Claude desktop app with no terminal at all, as an extension in code editors like VS Code and Cursor, and in the browser at claude.ai/code. All run the same engine, on Mac, Windows and Linux.
Is Claude only for coding?
No, on both counts. Claude the chat app writes, researches, analyzes and makes files. Claude Code is built for software, but Anthropic’s docs say it “can help with anything you can do from the command line,” and their own first-time examples include renaming the screenshots on your Desktop by what is in each image. If a task is about files on your computer, it can do it.
The honest framing: Claude Code is Claude with hands, pointed at a folder. Coding is what the hands are best at, because edit, run, check is how software gets built. But the folder can hold anything.
Who it is for
It is for the person who has outgrown a browser builder and wants the files. You do not need to read code. Anthropic’s guide for people who have never opened a terminal says it plainly: “You don’t need to know how to code. Describe what you want in plain English.” What you do need is to say what should happen, to open the app and check whether it did, and to keep a version you can return to.
You can learn more, and get the full prompts and the decision tables, in our course: Lesson 1, How to Direct Your AI Builder. The first lesson is free, no card.
In the next ten minutes, decide which question you were really asking. If it was “can Claude write code for me,” the chat app already does that, and you will hit the wall of placing and running the code yourself. If it was “can something build the app in my files while I direct it,” that is Claude Code, and the next step is a paid Claude plan and an empty folder on your Desktop.