Tell it how much you code with Claude and where Pro's limit stops you today. It names the cheapest plan that fits and prices the same usage at API rates.
Claude Code comes with Anthropic's Pro plan at $20 a month, and with its Max plans at $100 (5x) and $200 (20x). Max gives you five or twenty times Pro's usage per five-hour session. Anthropic publishes no fixed caps and no weekly multiple for Max, so most people guess.
The calculator works from what you already know. Set your days, hours and model mix, then say how far into a five-hour window Pro's limit stops you. You get a fit rating for each plan and an estimate of what the same work costs at API list price. Everything Anthropic publishes is linked below, checked in September 2026. Whatever Anthropic doesn't publish is labelled an estimate.
Opus 5.5 has been Claude Code's default model on Pro and Max since 2026-09-22, so 100% Opus is what you get if you never run /model.
Anthropic publishes no figure for where Pro stops you, so this is your own reading, and until you choose one the plans are not rated. Use a reading from after 2026-09-22, when Anthropic raised five-hour limits.
Anthropic does not publish Pro's cap, so tell the calculator where Pro stops you to rate each plan. At API list price this schedule is an estimated $376/month.
Every number here is an estimate, not a quote. Anthropic states Max 5x and Max 20x as five and twenty times Pro's usage per five-hour session and publishes no caps, no weekly multiple and no size for Pro's allowance, so the ratings scale your own Pro reading by those multiples and never rate the weekly limit. The API figure is list price for your schedule, built from Anthropic's published rates and two assumptions of ours: about 50K output tokens in a normal hour (100K in a heavy one), and 98% of input read from the prompt cache. Method, sources and worked examples are below. Prices and plan facts as of September 2026, from Anthropic's own pages.
Whichever plan you pick, it only pays off while the agent is running. Put the same Claude login on an always-on computer, start long runs inside tmux or from Telegram, and the hours you pay for keep working after you close the laptop.
Run Claude Code on HivraPro is $20 a month billed monthly, or $17 a month on the annual plan ($200 billed up front). Max 5x is $100 a month and Max 20x is $200, both billed monthly only. Prices exclude tax, and Anthropic says prices and plans are subject to change at its discretion. Claude Code is included in Pro and Max. Sources: Claude pricing, What is the Max plan?, What is the Pro plan?.
Anthropic states the Max multiples per five-hour session: Max 5x is five times Pro's per-session usage allowance, and Max 20x is twenty times. No Anthropic page states a weekly multiple, a token or message count, or the size of Pro's allowance, so the calculator prints none. To rate a plan, it takes the hours into a window where Pro stops you, multiplies them by 1, 5 or 20, and compares the result with the hours you work in your busiest five-hour window. Headroom means 60% of that capacity or less, tight means up to 100%, and over means more. The scaling is our arithmetic on the published multiples, and Anthropic doesn't promise it. The weekly limit is never rated. Sources: What is the Max plan?, Claude pricing.
Anthropic changes Claude Code's limits often. Five-hour limits doubled on 2026-05-06 and rose again on 2026-09-22. Weekly limits ran 50% higher from 2026-05-13 to 2026-09-13, and since 2026-09-14 they've been 25% higher than before that promotion. Anthropic didn't publish how big the 2026-09-22 increase was in hours or tokens. Any reading of where Pro stops you taken before 2026-09-22 is out of date, which is why the calculator asks for a recent one. Sources: Higher Claude Code limits announcement, Claude Code May to August 2026 weekly limits promotion, Claude Opus 5.5 announcement.
The dollar figure is an estimate at API list price. Don't read it as what a plan costs, or as a quote. Per million tokens, Opus 5.5 is $4 input, $20 output, $0.20 cache read and $5 for a five-minute cache write. Sonnet 5.5 is $2, $10, $0.20 and $2.50. Anthropic reports that Claude Code reads 324 input tokens for every output token, and that cache reads make up most of the cost. So the estimate counts 324 times the output tokens as input, prices 98% of that as cache reads and the rest as cache writes. That split is Hivra's assumption, and so are the 50K output tokens in a normal hour and 100K in a heavy one. Anthropic publishes none of them. Sources: Claude API pricing, Claude Opus 5.5 and sessions that use more context.
Anthropic's docs give a way to check those assumptions. They put API-billed Claude Code at about $13 per developer per active day, and $150 to $250 per developer per month across enterprise deployments. For 90% of users, costs stay below $30 per active day. We set the assumptions so that a normal three-hour day on Sonnet 5.5 comes to about $13.46, close to that average, and a heavy three-hour day on Opus 5.5 lands above the $30 line. The second figure on the page is the $13 enterprise average times your active days. It's an average, and it can't predict your own cost. Sources: Manage costs effectively (Claude Code docs).
Opus 5.5 has been Claude Code's default model on Pro, Max, Team and Enterprise since 2026-09-22, so the model mix starts at 100% Opus. Fable 5.1 ($10 input, $50 output) is never the default. Max includes it up to 50% of weekly limits and Pro uses usage credits for it, so the calculator leaves it out. It also leaves out Haiku 4.5 ($1 and $5), fast mode (available through usage credits only), Team and Enterprise seats, and usage credits themselves. Sources: Model configuration (Claude Code docs), Claude Fable models on your plan, Fast mode (Claude Code docs).
How this was checked: every value on this page was read from the Anthropic pages linked above in September 2026, then checked against them a second time. Anthropic changes plans often, so the month matters. Your inputs stay in your browser.
Last checked September 2026.
Inputs: 5 days a week, 3 hours a day, 100% Opus 5.5, normal use, Pro stops you about 2 hours in, weekly limit not hit.
Pro $20/mo Would hit limits Max 5x $100/mo Fits with headroom Max 20x $200/mo Fits with headroom API list price, estimated: $376/mo
Pro stops you two hours into a three-hour day. Five times two hours covers the window, so Max 5x is the cheapest plan that fits. The same work at API list price comes to an estimated $376 a month, well above Max 5x's $100.
Inputs: 1 day a week, 1 hour a day, 100% Sonnet 5.5, normal use, Pro never stops you, weekly limit not hit.
Pro $20/mo Fits with headroom Max 5x $100/mo Fits with headroom Max 20x $200/mo Fits with headroom API list price, estimated: $19.42/mo
Pro covers you. But one hour a week is light enough that the estimate at API list price, $19.42 a month, comes out a little under Pro's $20. That's the one kind of schedule where paying per token can be cheaper, and it leans on the assumptions above.
Inputs: 4 days a week, 6 hours a day, 50% Opus 5.5, normal use, Pro stops you about 1 hour in, Pro's weekly limit stops you.
Pro $20/mo Would hit limits Max 5x $100/mo Tight Max 20x $200/mo Fits with headroom API list price, estimated: $534/mo
Five times one hour is five hours against a five-hour window, so Max 5x is rated tight and Max 20x has headroom. The weekly stop on Pro is a separate limit, and the calculator can't rate it because Anthropic publishes no weekly multiple for Max. The verdict says so.
Inputs: 5 days a week, 5 hours a day, 100% Opus 5.5, normal use, Pro stops you about 30 minutes in, weekly limit not hit.
Pro $20/mo Would hit limits Max 5x $100/mo Would hit limits Max 20x $200/mo Fits with headroom API list price, estimated: $627/mo
Five times thirty minutes is about two and a half hours. That falls short of a five-hour window, so only Max 20x clears it. At API list price the same work comes to an estimated $627 a month, above Max 20x's $200.
Hivra, "Claude Code Plan Calculator", https://hivra.cloud/tools/claude-code-plan-calculator, facts last checked September 2026.
Source: <a href="https://hivra.cloud/tools/claude-code-plan-calculator">Hivra</a>, Claude Code Plan Calculator, facts last checked September 2026.
It takes the hours into a five-hour window where Pro stops you and multiplies them by the plan's published multiple (1, 5 or 20). Then it compares that with the hours you work in your busiest five-hour window. "Fits with headroom" means that window uses 60% of the capacity or less. "Tight" means up to 100%, and "Would hit limits" means more. The multiples come from Anthropic and the 60% line is ours. The rest is arithmetic, and the ratings promise nothing.
As of September 2026, no Anthropic page states a weekly multiple for Max 5x or Max 20x, so any weekly rating would be made up. If Pro's weekly limit is what stops you, tell the calculator. It adds a note that the ratings cover your five-hour windows only, and sends you to /usage after your first full week on a Max plan.
Anthropic doesn't publish how many hours or messages Pro covers, so there's no published baseline. Your own reading is the one real number in the calculation, and the published multiples scale it up to Max. Until you pick one, the plans show as not rated and only the API estimate appears. Use a reading from after 2026-09-22, when Anthropic raised five-hour limits. An older one is out of date.
It's your active hours times a cost per hour at API list price. Don't read it as a plan price or a quote. The cost per hour uses Anthropic's published rates and its published 324 input tokens per output token. Two assumptions are ours: about 50K output tokens in a normal hour (100K in a heavy one) and 98% of input read from the prompt cache. Anthropic publishes neither. A normal three-hour Sonnet 5.5 day comes to about $13.46, close to Anthropic's $13 enterprise average. Take it as a rough guide. It isn't a forecast for you.
Opus 5.5 has been Claude Code's default model on Pro and Max since 2026-09-22, so anyone who never runs /model is on Opus. If you switch, move the slider toward Sonnet 5.5. At list price it costs $2 against $4 per million input tokens and $10 against $20 per million output tokens. The calculator leaves the Fable models out.
It doesn't model Team and Enterprise seats, usage credits, fast mode, the Fable models or Haiku, or anything Anthropic hasn't published. What it covers is the three individual plans and API list price. For what each plan includes, what to do at a limit and when API billing is cheaper, read the plan pricing post linked below.
No. /usage shows your real plan usage and when it resets. Settings, then Usage, on your Claude account shows the session and weekly limits. The calculator estimates before you pay, from your own reading of where Pro stops you. Once you're on a plan, trust /usage over any estimate.
From $9.99/mo for 2 vCPU and 4 GB, not paused for inactivity. Your own Claude or ChatGPT login, or your own AI key. 7-day money-back guarantee on card payments.
Hivra is independent and is not affiliated with Anthropic.
