The full guide
Pricing, step by step
Claude Code comes with every paid Claude plan: Pro at $20 a month, Max at $100 or $200, and Team and Enterprise seats. Or you skip the plan and pay per token through the API. Which is cheaper depends on how much you use it — a number you can read off your own machine.
Claude Code plans and prices
The table at the top of this page is every plan as claude.com/pricing listed it on 3 October 2026. Three things it cannot hold:
- The pricing page says Max is "from $100". The $200 for Max 20× is in the help centre's Max plan article.
- Annual billing exists for Pro and Team only. Max is monthly; Enterprise is annual only, with usage on top at API rates.
- Prices exclude tax and are shown in local currency where Anthropic supports it.
Usage limits on Pro and Max
A flat plan is not unlimited: an allowance that resets every five hours, and a weekly limit on top, shared by Claude Code, claude.ai and Claude Desktop. How both windows work, and how close you are: Claude usage limits.
The API-key trap
With ANTHROPIC_API_KEY set, Claude Code bills that key instead of your plan once you approve it — and always in -p mode. You pay per token on top of the plan you already have. /status shows which one is in use. Setting one up, and when it is cheaper: Claude Code API key.
Claude Code API pricing, per token
Without a plan — an API key from the Claude Console or Microsoft Foundry — every token is billed at the rates in the table above. Amazon Bedrock and Google Cloud's Agent Platform (formerly Vertex AI) set their own Claude prices, and their regional and multi-region endpoints cost 10% more than their global ones. Claude 4.6 and later models, except Haiku 5.5, run their full 1M-token context at the standard price.
For Claude Code the line that matters is cache read. Every turn re-sends the whole conversation, almost all of it from the cache, so cache reads were 78% to 83% of each month's list price on my machine. Where the money goes breaks a bill down by what the turns did, and Claude API pricing has every model, the modifiers and a per-call calculator.
How much does Claude Code cost per month?
Anthropic's average is $150–250 per developer a month; my median day came to $153 at API list prices, which is one heavy user, not a forecast.
Anthropic's cost page takes that average across enterprise deployments: around $13 per developer per active day, with 90% of users under $30 a day.
My own machine, priced at API list prices: the 56 days I used Claude Code from 18 July to 2 October 2026, by UTC date. The middle half of my days ran from $75 to $373, and August came to $5,898 on a $200 plan.
That median is about 12 times Anthropic's average day, from one developer who works in it all day; the month-by-month post shows how it was read.
Pro, Max or the API: which one you need
Two questions decide it.
- Do you hit the limit? If you keep waiting for the five-hour or the weekly reset, Max 5× gives five times Pro's allowance per session, and Max 20× twenty times.
- What is your use worth at list price? If a month of your tokens at API prices comes to less than the plan fee, the API is cheaper. If it comes to more, the plan is — my August was 29 times the $200 fee.
A team that wants one bill picks Team seats; every seat runs Claude Code, and which seat a developer needs is a question of the window: Claude Team plan. On Enterprise, usage is billed at API rates, so the list price is roughly the bill before any discount negotiated with Anthropic. What Anthropic publishes about the Enterprise seat and its usage line, and how to size that line from cost per merged pull request: Claude Enterprise pricing.
Running Codex too? Codex vs Claude Code puts the two side by side, with what each cost on one machine.
Claude Max 20× plan
Max 20× is the largest individual plan: $200 a month, billed monthly only, for twenty times Pro's usage per session. From Anthropic's help centre, read on 8 October 2026:
- Limits: the session allowance resets every five hours, and a weekly limit applies across all models, resetting at a fixed time assigned to your account.
- Upgrading from Max 5×: charged pro rata for the rest of the billing cycle.
- API credits: $200 a month for the Claude API, claimed by linking a Claude Console organization. They cover the API and the Agent SDK, not interactive Claude Code or extra usage, and expire at the end of each cycle.
- Discounts: Anthropic offers no standing discount on Max.
On my machine Max 20× was never near either cap: the highest five-hour window read 17% and the highest week 27%, on the days its status line was installed. Every five-hour and weekly reading: Claude usage limits.
Is Claude Max worth it?
For me, yes. On the $200 plan, August's tokens were worth $5,898 at list price, and from 20 August to the end of September Claude Code did not stop me once for a usage limit. That window holds 23 days of Claude Code, not 42: from 10 to 28 September my work went to Codex. I have since moved to the $100 plan. That is one heavy user, not a rule.
If your month at list price is under $100, it is not: Pro or the API costs less. The way to know is to read that number, not guess it.
See what your own use costs
Inside Claude Code, /usage (also /cost and /stats) shows the session's estimated cost if you pay by token, your plan's usage bars if you subscribe, and what has used them over the last day or week. It cannot go back further than the transcripts on disk, and Claude Code deletes those after 30 days.
Caprock, a free desktop app, records every session on the machine, whether or not it started it, and prices it at the list prices above — an estimate, as Claude Code's own /usage is, beside the plan's dollars and the share of the weekly limit the work used. Set your plan in the header and the Cost screen puts the selected range's list price — 30 days by default — against what you pay for it. With its status line installed (caprock statusline install), it also shows how much of the five-hour and weekly windows is used and when each resets, on Pro and Max.
Free, open source; nothing leaves your machine. Install Caprock and it reads back what Claude Code still has on disk.
Questions people ask
- Is Claude Code free?
- No. The free Claude plan does not include Claude Code; the cheapest way in is Pro at $20 a month, or paying per token with an API key.
- What is the difference between Claude Pro and Max?
- Usage. Pro is $20 a month; Max 5× is $100 for five times Pro's usage per session and Max 20× is $200 for twenty times. Both include Claude Code.
- What does the Claude Max 20× plan include?
- Twenty times Pro's usage per session for $200 a month, billed monthly, with a five-hour session limit and a weekly limit across all models. It also includes $200 a month of Claude API credits, which do not cover interactive Claude Code.
- How much is Claude Pro billed annually?
- $17 a month, $200 billed up front, against $20 billed monthly. Max has no annual price; Team does, at $20 a standard seat.
- How much does Claude Team cost?
- A standard seat is $25 a month, or $20 billed annually; a premium seat with five times the usage is $125, or $100 annually. Both include Claude Code.
- Does Claude Code share usage limits with claude.ai?
- Yes. Claude Code, claude.ai and Claude Desktop draw from one allowance, so a long chat in the browser leaves less for code in the terminal.
- How much does Claude Code cost per month?
- Anthropic's average across enterprise deployments is $150–250 per developer a month. My median day came to $153 at API list prices: one heavy user, not a forecast.
- Is Claude Max worth it?
- If a month of your tokens at list price is over $100 and Pro's limits stop you, yes. My August on Max 20× was worth $5,898 at list price for a $200 fee.
Sources
- Claude plans and prices, read 3 October 2026
- Claude Help Center — What is the Max plan?, read 8 October 2026
- Claude Help Center — Monthly API credits for Max and Team plans, read 8 October 2026
- Claude Help Center — What is the Pro plan?
- Claude Help Center — How do usage and length limits work?
- Claude Help Center — Using Claude Code with your Pro or Max plan
- Claude Help Center — Usage credits
- Anthropic API pricing, read 3 October 2026
- Claude Code docs — Manage costs effectively, read 20 October 2026
- Claude Code docs — Commands (/usage, /cost, /stats)
- Caprock's dated price table
- Caprock UI spec — the Cost screen, Plan value and Plan limits
Try it
The record of every coding agent you run, the sessions it did not start included. Free, open source; nothing leaves your machine.
Notarized by Apple. One file for Apple silicon and Intel, macOS 13 or later.
Or run it yourself:
Paste in Terminal: it installs Caprock and opens it.Or with Homebrew: brew install --cask dspv/tap/caprock-app