How much does Claude Code cost: $20 to start, $200 at the top

How much does Claude Code cost? The cheapest steady way in is Claude Pro at $20 a month, and the individual plans top out at $200 a month on Ultra. Teams pay per seat, companies pay per seat plus tokens, and anyone can skip plans and pay per token with an API key.

So the range is wide, and both ends are real: a $20 plan that stops you at a limit, and a token bill with no ceiling unless you set one. Here is the price list as of November 2026, from Anthropic's pricing page.

The price list, one line each

• Free plan: $0, and Claude Code is not part of it.
• Pro: $20 a month, or $17 a month on a yearly plan.
• Max: $100 a month, five times Pro's usage per five-hour window and per week.
• Ultra: $200 a month, ten times Pro per week, and no five-hour limit once Anthropic verifies the account.
• Team: $25 a seat a month for a standard seat, $125 for a premium seat, less on yearly billing.
• Enterprise: $20 a seat a month on an annual contract with at least 20 seats, plus all usage at API rates.
• API key: no subscription, every token billed. Sonnet 5.5 costs $2 per million input tokens and $10 per million output tokens.

The installer itself is free. Running it is not.

Why people still say "Max 20x"

The $200 plan used to be called Max 20x. Anthropic renamed it Ultra and moved existing subscribers over at the same price. Its limit changed shape too: Ultra counts usage per week, and verified accounts lose the five-hour cap. If a thread from earlier this year quotes "20x Pro per session", it describes a plan that no longer goes by that name.

The cost that isn't a cost

Run /usage in Claude Code on a Pro or Max plan and you will see a session total in dollars. Do not panic. Anthropic's docs say that figure is meant for API users, and subscribers have the usage included in the plan. It shows what the session would cost at list price, which makes it useful for comparison, but you are not charged for it.

The setting that does cost money

Two things can put a real charge on top of a subscription, and the first is easy to miss. If ANTHROPIC_API_KEY is set in your shell, Claude Code uses that key instead of your plan, according to Anthropic's help center. Every token then goes on the API account. If a Console balance drops while you thought you were on a plan, check your environment variables first.

The second is usage credits, which you have to turn on yourself. They let a paid plan keep working past a limit at API rates, billed separately.

How to tell which tier you need

Upgrading on a hunch is how people overpay. Your own logs give a better answer.

UsageAtlas is a free, open-source desktop app that reads Claude Code's local logs and your plan limits, then shows tokens by model, an estimated cost at API prices and how close you are to each limit. It keeps that history after Claude Code deletes its local files at 30 days, so you can look back over a full month before you upgrade. No account, no prompts read, and it runs on Windows, macOS and Linux.

The dollar figures in any tracker, including Claude Code's own, are estimates. Anthropic's invoice is the bill.

When the answer is "it depends on your employer"

If your company already pays for Claude Team or Enterprise, your Claude Code cost is somebody else's line item. Ask your admin about your seat type before you buy a personal plan on top.