How Much Does Claude Code Cost - Plans, Pricing, and Which One a Business Needs

How much does Claude Code cost?

Claude Code costs either a flat monthly subscription or a per-use API rate, and you choose which. The subscription tiers are Pro at $20 a month, Max at $100 a month, and a larger Max at $200 a month, each billed month to month with a small discount for paying annually. The subscription bundles Claude Code together with the Claude desktop and web apps under one shared usage allowance. The alternative is the API, where you skip the subscription entirely and pay per token of input and output, with no monthly fee and no usage cap.

That is the whole pricing picture, and the confusion around it comes from treating those as two prices for one thing. They are two different billing models. A subscription is a fixed bill with a usage ceiling. The API is a variable bill with no ceiling. Neither is cheaper in the abstract. Which one costs you less depends entirely on how much you actually run, which is why the honest answer to the cost question starts with a plan you can measure against.

What do the subscription plans include?

Each subscription tier buys you more of the same thing: usage. Pro at $20 gives a modest allowance suited to light and occasional work. Max at $100 raises that allowance several times over, and the $200 Max raises it again for people who run Claude Code for hours every day. On all of these, Claude Code and the Claude apps draw from one shared pool, so heavy chatting in the app eats into the same budget your building does.

The number that decides your plan is not the price, it is the usage limit attached to it. Subscriptions run on a session window plus a longer weekly cap, and when you hit the cap you wait or upgrade. For a business that matters more than the dollar figure, because a plan you constantly hit is a plan that keeps interrupting real work. Start low, watch whether you hit the ceiling, and move up only when you do.

When does the pay-per-use API cost less than a subscription?

The API costs less when your usage is genuinely light or genuinely spiky. If you open Claude Code a few times a week for short jobs, paying per token can come in under $20 a month, because you are only billed for what you touch. It also has no cap, so a sudden heavy week does not lock you out the way a subscription limit does. You pay for that week and move on.

The trade is predictability. A subscription is a known number on your card every month. The API is a variable bill that climbs with use and has no built-in ceiling to stop it, which is exactly what makes it dangerous for a team that has not set a limit. Anyone routing Claude Code through the API for a business should set a hard spend cap in the billing console first, before the first heavy day arrives rather than after it.

What is the Cost Fit Rule for choosing a plan?

The Cost Fit Rule is our way of picking the cheapest option that still covers your work: start on the lowest plan, run your real jobs for two weeks, and only move up when you actually hit the limit. Most owners overbuy on day one because the bigger plan feels safer. It is not safer, it is just more expensive, because a plan you never max out is money spent on headroom you did not need.

1. Start on Pro at $20. It is the cheapest way to learn whether Claude Code fits how you work at all.
2. Use it on real jobs for two weeks, not test prompts. The fit question is about your actual workload, not a demo.
3. Watch for the usage limit. If you never hit it, you are on the right plan and you are done.
4. Only upgrade when the limit interrupts real work more than once. That is the signal you have outgrown the tier, not a hunch.
5. Consider the API only if your use is light and occasional, and only after you set a spend cap.

The rule works because the cost of being wrong is small in one direction and large in the other. Start too low and the fix is a one-click upgrade the day you hit the ceiling. Start too high and you quietly overpay every month until someone questions the line item. The cheap mistake is the reversible one, so make that mistake on purpose.

Does the subscription include the desktop app and the Claude apps?

Yes. A Pro or Max subscription covers Claude Code, the Claude desktop app, and the web app together, all drawing from one shared usage allowance. That is the point worth internalising for a business, because it means you are not buying Claude Code as a separate line item on top of a Claude subscription you already pay for. One subscription, one bill, one pool of usage across all of it.

This is also why the desktop app is the sensible front door rather than the terminal. If you already pay for Claude, you already have Claude Code, and you can use it from the desktop app without touching a command line. [How to use Claude Code for your business](/blog/how-to-use-claude-code-for-your-business) covers getting started that way, and none of it requires a developer setup or an extra charge.

What are the hidden costs people forget?

The subscription is rarely the real cost. The real cost is the hour a week someone spends learning to brief the tool well, the occasional cloud or hosting bill for something you build with it, and the time to check its output before you rely on it. None of those show up on the Anthropic invoice, and all of them are larger than $20 for a business that is serious about the work.

• Learning time. The first two weeks are slower before they are faster, and that time is a real cost you should expect.
• Downstream bills. Something Claude Code builds may need hosting, a database, or a paid connector, none of which are in the subscription.
• Verification time. Output has to be checked before it goes to a customer, and that habit is a cost you should build in, not skip.
• The overbuy tax. A Max plan you never fill is the most common hidden cost of all, and it is entirely self-inflicted.

If you are weighing Claude Code against paying a person to do the same job, the comparison is not the subscription against a salary. It is the subscription plus your learning time against the salary. [Claude Code vs hiring a developer](/blog/claude-code-vs-hiring-a-developer) works through that honestly, including the jobs where hiring is still the right call.

Is Claude Code worth the cost for a small business?

For most small businesses, $20 a month is worth it if you have even one recurring job it can take off your plate. The maths is not complicated: if one plan replaces a few hours of work a month, or one tool subscription you were paying for a job Claude Code now does, it has paid for itself. The risk is not the $20, it is spending the $20 and never actually using it.

The way to make the cost worth it is to point it at a specific job in the first week rather than exploring it in general. Build one small internal tool, automate one report, replace one piece of software you were renting. [How to build an internal tool with Claude Code instead of buying software](/blog/how-to-build-an-internal-tool-with-claude-code-instead-of-buying-software) is a good first target, because the saved subscription often covers the Claude Code plan on its own.