No. There is no free version of Claude Code, and no free trial of it either. It comes with a paid Claude plan, and Anthropic’s setup page says plainly that the free plan does not include it. Sign in to the app or the terminal with a free account and you are asked to upgrade. The longer answer is that “what it costs” has a different shape from the builders you are used to. There are no credits, no per-prompt meter, and nothing that runs out in the middle of a fix. There is one trap that makes it cost more than your subscription, and one no-card way to try this way of working before you pay anyone.
What the paid plan actually buys
Claude Code is included in the paid Claude plans, the same subscription that pays for the chat app. Anthropic’s pricing page lists it as a feature of the paid individual plans, and the help center says you log in with the same credentials you use for Claude. There is no separate product to buy, no add-on, and no key to paste. One subscription covers Claude Code in the desktop app, the terminal, a code editor and the browser, plus the chat app.
That is the first thing that feels strange coming from Lovable or Base44: a flat monthly fee for a tool that then works as much as the plan allows, instead of a balance that goes down every time you ask for something.
It is metered by time, not by credits
On a Claude plan, usage is a budget that refills on a clock. Anthropic’s help center describes two limits. A session limit resets every five hours. A weekly limit sits on top and resets at a fixed time each week assigned to your account. Both are shared across everything you do with Claude: the chat app, Claude Code in the terminal, and the desktop app all draw from the same allowance. You can see both as progress bars under Settings, then Usage, in your Claude account.
What uses the budget faster is what uses any AI faster: long conversations, big files, and the larger models. A long afternoon of building can hit the session limit. When it does, nothing is charged. You wait for the window to reset, and the usage page shows when. If you would rather not wait, paid plans let you switch on usage credits, an opt-in billed at Anthropic’s API rates that comes with a monthly cap you set yourself. Nothing is bought unless you turn it on.
Compare that with a credit-metered builder, where the fifth attempt at the same bug costs as much as the first and the balance is gone before the app works. Here, a bad afternoon costs you a wait. That changes how you work more than any feature does.
The trap that makes it cost more than your subscription
An API key is a password-like code that lets a program use Claude’s pay-per-use service, where you are billed by the amount of text in and out, with no cap. If you have ever pasted one into a tool on this computer, it may still be saved there as a setting called ANTHROPIC_API_KEY. Anthropic’s help center is explicit about what happens next: if that setting is present, Claude Code in the terminal uses the key instead of your subscription, and you get API charges on top of the plan you already pay for.
This answers the “is an API key free” question too. No. A key is the pay-per-use door, separate from any chat subscription, and it never runs out on its own. It keeps billing. Claude Code asks you once, at first start, whether to use a key it finds. Decline. Typing /status inside Claude Code shows an API key row only when a key is in use. The desktop app does not read that setting, one more reason a non-coder should start there.
What the subscription does not pay for
Claude Code writes and edits files on your computer. It does not run your app for the world, and the plan does not include the things a live app needs. Hosting, the service that keeps your app reachable at a web address, is a separate account. So is a database, so is a login service, and so is a custom domain. Each has its own free tier or bill, and each is a choice you make later. Lovable bundled all of that into one credit balance, which is convenient right up to the moment you want to change one of them.
The other thing it does not pay for is your app’s own AI. If your app answers users with AI, that is a separate service your app talks to while it runs, billed pay-per-use through an API key of its own. That key belongs in your app’s settings, never on your computer as the setting above, and never in a page where a visitor can read it. Two different AIs, two different bills: the builder on a subscription, the brain inside your app by usage.
The no-card way to try this way of working
If you want to try directing a builder in plain language before paying anyone, use Codex. It is OpenAI’s tool that does the same job as Claude Code, and OpenAI’s pricing page lists it as included in ChatGPT’s free plan, with the free tier described as enough to explore it on quick coding tasks. That is honest sizing: a free allowance gets you through a first small build, not through weeks of work, and at some point you will pay for whichever builder you settle on. But you never need a card to begin, and the walls you hit are the same either way.
What does not exist is a legitimate free Claude Code. Anything promising a free version, a cracked build or a shared account is offering someone else’s login, or worse. The tool signs in through your Claude account, and that account decides what you can use.
You can learn more, and get the full prompts and the decision tables, in our course, starting from You’ve Outgrown Lovable. The first lesson is free, no card.
In the next ten minutes: open your Claude account settings and check your plan and the usage page. If you are on the free plan and not ready to pay, sign in to ChatGPT and try Codex on something small. If you are already paying, you already own Claude Code. Open it and ask for one change you can see.