How much is Claude Code?
Claude Code has no separate price — it's included in a Claude subscription, so the cheapest way in is Claude Pro at $20/month ($17 billed annually). Heavy users move to Max at $100 or $200. The alternative is Console credits: no subscription, you pay per token consumed, which is cheaper for light use and unbounded for heavy use.
- Claude Pro
- $20 / mo · $17 annually
- Max 5×
- $100 / mo · 5× Pro usage
- Max 20×
- $200 / mo · monthly billing only
- Console credits
- Per token no subscription
- Free tier
- None install is free, use isn't
- Team / Enterprise
- Per seat quoted by Anthropic
- Separate CC fee
- $0 included in the plan
- Prices verified
- 2026-07-16 we re-check quarterly
Two billing routes, and the one number that isn't a number.
The subscription figures above come from our verified plan sheet, checked 2026-07-16 against Anthropic's pricing page — the same sheet behind the subscription-vs-API calculator, so a Q&A page and a calculator can't quietly disagree.
The second route has no headline price at all, and that's the honest answer rather than a dodge. On Console credits you pay for tokens consumed, so the bill is a function of how much code the agent reads and writes — a quiet week and a large refactor differ by an order of magnitude. Current per-token rates for every Claude model live on our Anthropic pricing page, derived from the snapshot rather than retyped here, so they stay right when Anthropic reprices.
Team and Enterprise are sold per seat and quoted by Anthropic. They sit outside our verified sheet, so we describe them rather than put a number on them.
What actually drives the bill.
On Pro or Max you pay a flat monthly fee and work within usage limits. The appeal is a predictable invoice; the friction is that limits bite mid-task rather than at a line item you can inspect. Max exists precisely because agentic coding burns through far more of the allowance than chat does — an agent reading a repository consumes tokens at a rate a conversation never approaches.
If you're weighing tiers rather than routes, we've compared Free vs Pro and Pro vs Max separately.
Console credits price the same work per token: input, output, and cache reads and writes. That is excellent when you use the tool in bursts — you pay for the burst and nothing in the quiet weeks — and merciless during sustained agentic work, because there's no ceiling.
Two levers matter more than the headline rate. Prompt caching cuts the cost of context you send repeatedly, which is most of what an agent does with a codebase. And model choice swings the bill several-fold: the same task on a smaller model costs a fraction of the flagship. Rates for each are on the Anthropic pricing page.
There's no universal answer, and anyone giving you one hasn't asked how much you code. The break-even sits where your monthly token spend crosses the subscription fee — below it, credits win; above it, the flat fee does, and Max exists for the tail beyond that.
Rather than guess, put your own numbers in: our subscription-vs-API calculator takes your volume and returns the crossover for the current rates. If you're already on credits, checking your usage inside the CLI is the fastest way to see where you actually land.
If the number is too high.
Two directions that genuinely lower the bill, rather than the proxy tricks that swap Claude for a different model without saying so.
Frequently asked.
Quick follow-ups people search after this question.