How to Subscribe to Claude Pro — Choosing Between Pro and Max, Payment Setup, and Canceling

The question a Claude subscription guide should answer first is which tier to buy, not how to click the subscribe button. Claude currently sells three individual tiers: Pro at $20/month, Max 5x at about $100/month, Max 20x at about $200/month (all official list prices; what actually gets charged also depends on the Tax line at checkout). This guide works through the usage gap between the three tiers, who each one fits, and the signals that mean it is time to move up, then covers the parts of paying and adding a card that actually matter.

Why put the weight on picking a tier rather than on payment? Because once you look at the payment side in two layers, it turns out far cleaner than most people expect.

Put another way: get the BIN range right and adding a card is not hard. What is hard is picking the wrong tier — either you pay $200 and never use it up, or you sit on the $20 tier hitting the ceiling every day. So this guide covers tier selection first, payment second.

1. The 30-Second Answer: Which Claude Tier to Buy

• You hit the usage ceiling less than once a week → stay on Pro ($20/month), and do not pay extra for usage you might need someday.
• You hit the ceiling 2–3 times a week and wait hours each time → move to Max 5x (about $100/month).
• You hit it more than once a day, or you run Claude Code against real repositories → go straight to Max 20x (about $200/month). By the forced-downtime math in section 3, 20x usually beats 5x at that level of use.
• Work the payment side in two layers: first confirm the BIN range supports AI subscription scenarios (layer one — the BIN range notes on the card-issuing page say so); once the range is right, adding the card is a $0 verification authorization that passes if you copy the details correctly, and what actually fails is the real charge afterward, plus merchant-side upfront rejections caused by a mismatch between the country of the card, the country of the proxy IP, and the country of the billing address (layer two).
• ⛔ The most expensive move after a decline is hammering retry: fix the cause, wait at least an hour, and never make a third consecutive attempt. Repeated failures in a short window usually trip the payment gateway anti-abuse protection, after which even a spotless setup gets declined instantly. Details in section 7.

2. Claude Subscription Pricing: Pro / Max 5x / Max 20x Compared

Conclusion first, before you choose a plan: what separates the three tiers is mostly the usage allowance, not the feature list. The 5x and 20x in the tier names are multiples of Pro usage, not a signal that some new model has been unlocked. Choosing a tier is fundamentally about buying time, not about buying features.

There is also a Team tier aimed at organizations (billed per seat, with a minimum seat count), which individual users cannot use. The figures here are official list prices, before tax; the real price is whatever your checkout page shows.

What Max 5x and 20x have in common (Pro needs a separate check)

The two Max tiers are identical in available models, context length, Artifacts, Projects, and Claude Code; the only difference is the allowance. Pro needs to be checked on its own: the model tiers Claude Code can call under Pro and under Max are not entirely the same, and some new features reach tiers at different times. Which features are gated by tier changes over time, so before you buy, go by your checkout page and the current plan description on the official site.

3. Max 5x vs Max 20x: Decide by Forced Downtime, Not by Feel

Conclusion first: if hitting the usage ceiling forces more than 2 hours of downtime a week, Max 20x is already the better deal than Max 5x, even though 5x looks half the price. The reason is that a subscription fee competes with your working hours, not with your wallet.

A calculation you can run yourself

1. Log one week first: every time a usage notice appears, write down the time, how long you waited, and whether you switched to something else because of it.
2. Add up the waiting time. Count the time you were genuinely interrupted, not the reset window the product quotes — plenty of people go do something else after hitting the limit and come back half an hour after it cleared. That half hour counts too.
3. Divide the price gap by it. 20x costs about $100/month more than 5x; across 22 working days that is about $4.5 a day: if you lose more than half an hour a day to interruptions, the gap pays for itself; if you hit the limit once a week and only on a crunch day, what you need is not a higher tier but a different way of spreading out that one day.

If Claude Code is your main use case: choosing between the two Max plans

Claude Code on a real project consumes on a different scale from chatting: it reads the whole repository, runs tests, and edits files repeatedly, so one medium task is worth dozens of chat turns. For users whose main tool is Claude Code, Pro is barely enough to try it out, 5x covers small projects, and running a large repository long term needs 20x. For the full setup that gets Claude Code working (network and payment included), see getting Claude Code running from China.

4. Four Signals to Upgrade from Pro to Max (and Three Times Not To)

Conclusion: the trigger for upgrading should be frequency, not frustration. If any two of the four below apply, it is time to move up.

• Signal 1: you hit the limit ≥ 2 times a week, and not all piled into a single day.
• Signal 2: you have started degrading how you work to save allowance — not daring to paste in a whole file, feeding it a chunk at a time, deliberately asking fewer follow-ups. This is the most hidden cost: it shows up not as waiting but as worse output.
• Signal 3: you are already running two accounts in parallel for the same work, rotating between them. That breaks platform rules and risks a ban, and all-in it often costs more than simply moving up a tier (for the collateral damage, see why Claude / ChatGPT accounts get banned).
• Signal 4: Claude is now on your delivery path — not help me think this through, but this code ships today. Once output is tied to a deadline, running out of allowance stops being a matter of mood.

Three cases where you should hold off

1. You hit the limit because you keep pasting the same long document into the same conversation — open a new conversation, summarize once, then carry on. For a lot of people that single step ends the problem.
2. You only use it intensively on weekends — what you are hitting is the 5-hour window, not a monthly total, so spreading the work across two days is cheaper.
3. What you actually need is the API — running fixed tasks in bulk is usually cheaper on pay-per-use, and the subscription allowance and the API allowance are two entirely separate systems.

5. How to Pay for a Claude Subscription: Hard Requirements on BIN Range, Card, and Environment

This section follows the two layers in order: get the BIN range right first (layer one), then get the card and the environment right (layer two). Reversing that order carries a direct cost — on a BIN range whose issuer policy does not support AI subscriptions at all, a perfect score on every environment condition below buys you nothing.

Layer one: confirm the BIN range supports AI subscription scenarios

The issuers behind different BIN ranges hold different policies on merchant scenarios. Some ranges have an issuer policy that explicitly does not support scenarios such as AI subscriptions — this is not risk control finding you suspicious; that range simply does not do this line of business. Taking such a range to Claude fails, and the cause is an unsupported BIN range. Switching to a range that supports AI scenarios solves it; troubleshooting in place is wasted effort.

• Read the BIN range notes before you issue a card: the card-issuing page states which merchant scenarios that range serves. Pick AI subscription as the intended use, and the range you get is one whose issuer policy supports that kind of merchant.
• If you already issued the card and then find it is wrong: the card detail page also shows the usage policy for that range. If the range does not support AI subscriptions, stop working on your environment and issue a new card on a different range.
• One term that is easy to confuse: when this site says acceptance rate, it means the issuer scenario-support rate, not an authorization success rate — it answers whether that range serves this kind of merchant, not whether a given transaction will go through.
• For which scenarios each range supports, go by the BIN range detail page and what the card-issuing page currently shows, rather than from memory of this article or any older post.

Layer two: with the right range, get the card and the environment right

Once the range is right, keep two things apart or your troubleshooting will point the wrong way: card verification is a $0 authorization that passes as long as the card number, expiry date, and CVC are copied correctly; what actually fails is the real monthly charge afterward, plus merchant-side upfront rejections where the request never reached the card at all. The conditions below are ones our support team has confirmed over and over in real tickets, and following them clears the most common failure causes in advance.

Account side: do not add a card to an account created ten minutes ago

The account should be at least a week old, with normal usage during that week. Adding a card sooner — especially on an account freshly created through a proxy — fails often. Letting an account age before paying is not folklore: risk control looks at the behavior curve of the account, and one that spends $200 on the day it was created is an outlier sample in any payment risk model.

Network side: stable beats premium

• Your IP must be stable for the 24–48 hours before you add the card, with no jumps across countries or regions. If you switched nodes within those two days, let it settle for 24 hours before adding the card.
• The gap between datacenter IPs and residential IPs is smaller than folklore suggests; what matters is that it is yours alone — a public node shared by many people has likely been flagged long ago, and no amount of card hygiene rescues that.
• Keep that same route for the whole life of the subscription, not just at the moment of payment.

★ Three-way consistency: card, IP, billing address

This is the line every guide mentions in passing and almost nobody explains: the country of the card = the country of the proxy IP = the country of the billing address, and ideally the state of the IP = the state of the billing address. The part below is where the real information gap sits:

• Line 1 of the address takes the street only, and line 2 (Address line 2 / Apt, suite) stays empty — put nothing there. Plenty of failures come down to meaningless text typed into line 2.
• City, state, and ZIP must be consistent with one another. Inventing a ZIP on the spot is the most common avoidable failure.
• ⛔ Do not copy the universal address that circulates online. Thousands of people sharing one address string is itself a textbook risk signal — this matters far more than whether the address looks tidy.
• A word on tax while we are here: US state tax rates differ, so on the same tier the amount actually charged can vary by a dollar or two between people. That is a fact, but it is not worth reshuffling your address over — go by the Tax line at checkout and keep a few extra dollars of buffer when funding the card (see section 8).

Card side: enough limit, copy buttons, never retype

• Use a card on a US BIN, and keep the available limit above list price + tax + a little buffer (exact figures in section 8). On BIN ranges that support AI scenarios, the most frequent failure cause in our ledger is an insufficient per-transaction limit, particularly a $200 annual or Max charge meeting a low-limit card.
• Always use the copy buttons for card number, expiry date, and CVV. A mistyped CVC or a reversed expiry date has long been the second most frequent cause, and that kind of failure does not even clear the $0 verification.

6. The Claude Subscription Flow: 6 Steps

The flow itself is not complicated; the order matters more than the steps — put picking the right BIN range, aging the account, and stabilizing the IP up front, and the rest is mostly filling in a form.

1. Let the account sit for a week, and actually use it: after signing up, chat normally on the free allowance for a few days, and do not keep switching environments across several devices.
2. Fix your route and keep it stable for at least 24 hours: open any IP lookup page and check for yourself that the country reads US, the city is a real city, and it is in the same country as the billing address on the card (ideally the same state).
3. Issue a card on a BIN range that supports AI subscription scenarios, and fund the limit properly: when issuing, select AI subscription as the intended use, and the BIN range notes will state which scenarios it serves. Suggested funding: Pro ≥ $25 (the Claude ledger shows no tax spread and no supplementary charges, so $25 is enough), Max 5x ≥ $110, Max 20x ≥ $210. The subscription is charged once a month, so do not fund only the first month — running short next month means the subscription stops, not that you carry a debt. Minimum top-up amounts and minimum issuance limits are whatever the top-up page and card-issuing page show.
4. Pick the tier inside claude.ai: avatar in the bottom-left → Settings → Billing / Upgrade, where Pro and Max are listed side by side and Max splits into 5x and 20x. Go with the tier you settled on in sections 2 and 3, and do not change your mind on the checkout page — repeatedly entering and leaving checkout also counts as risk-control behavior.
5. Enter the card and the billing address: paste card number, expiry date, and CVV using the copy buttons; for the billing address, copy word for word the one shown on the card detail page, street only on line 1, line 2 empty, and keep the city / state / ZIP combination the card detail page gives you without changing any single part of it.
6. Check the Tax line before you submit: list price + Tax = the amount actually charged, so confirm your limit covers it before clicking confirm. If it fails, do not hammer the button — skip straight to the next section.

7. When the Card Is Declined: Check the BIN Range, Then Triage, Then Cool Down

Conclusion: on the first decline, the most expensive move is to click again immediately. Tracing back through our support tickets, a burst of attempts in a short window usually trips the brute-force protection on the payment gateway, the account or the card goes into an observation state, and after that you get declined instantly no matter how clean you make the environment — plenty of people conclude the card is bad and burn through several cards, when in fact they were locked out by their own rapid retries.

One thing to be clear about: the exact thresholds and cooldown durations of the payment gateway are external parameters that are never published; what follows are empirical values observed on the support side over a long period, not official figures. So treat the schedule below as better to wait longer, and do not use it to test the boundary.

Step 0: confirm this card sits on a BIN range that supports AI subscription scenarios

This step always comes first, because it decides whether any of the troubleshooting that follows means anything. If the issuer policy behind this card BIN range does not support merchant scenarios such as AI subscriptions, the failure is a case of an unsupported BIN range — move to a range that supports AI scenarios and it is solved; changing your IP, editing the address, or retrying a hundred times changes nothing. For which scenarios a range serves, read the BIN range notes on the card-issuing page or the usage policy on the card detail page. Once the range checks out, move on to the triage table.

Step 1: triage — did the failure happen on the card side or the merchant side

The no-record-at-all row is the one worth remembering: it means the rejection happened in the upfront risk checks of the merchant and the payment gateway, and the card was never asked. In that case a new card, a top-up, or a higher limit all do nothing; the fix has to come from the account and the network environment. For the full list of decline causes ranked by frequency, see the top 10 reasons a virtual card gets declined.

Step 2: run the cooldown policy by attempt count

We review failed tickets month by month: on BIN ranges that support AI scenarios, we see no stable pattern of an environment that is entirely correct yet gets declined over and over — the overwhelming majority of repeat failures trace back to an account that is too new, a jumping IP, a malformed address, an insufficient limit on the card, or an existing lock from rapid retries; while the rest of the repeat failures do not belong to this layer at all, and come from subscribing on a BIN range that does not support AI subscription scenarios.

8. Planning the Annual Cost and the Balance on the Card

Conclusion: look at this expense by the year, not by the month. Max 20x runs about $2,400 a year — a number that makes many people rethink whether they really use it every day, which is exactly the right question to sit with.

• Why $25 is enough for the $20 tier: checkout tax varies with the state of the billing address, so estimate the $20 monthly tier at about $22 including tax and keep $25 on the card, which leaves a comfortable $3 of buffer. The Claude ledger shows no tax spread and no supplementary charges (the charges observed are the two round values $20.00 and $200.00), so there is no need to add another step the way you would for ChatGPT. Same logic on the $200 tier: fund it at ≥ $210.
• Check the limit two days before the renewal date. A failed renewal is not carried as a debt; it drops you straight off the tier, and getting back on means going through the payment risk checks again.
• Factor in two costs: the USDT account top-up fee is 0.5% (about $1 on a $200 subscription), and top-ups are usually credited within minutes; the card top-up service fee starts at 2% and steps down with your calendar-month cumulative card top-up volume, to a floor of 0.5%, so year-round Max 20x users pay less the more they load. Card issuance starts at $1 each (it varies by BIN range — the shared-limit card recommended for AI subscriptions is $2, shown right on the card-issuing page). Minimum top-up amounts, minimum issuance limits, and per-card limit caps are all adjustable, so go by what the top-up page and card-issuing page currently show. For the full breakdown, see the fee schedule.
• If an annual payment option appears at checkout, the unit price is usually lower, but it is a single large charge and the card has to hold the whole year at once. If you also subscribe to ChatGPT, look at both bills first, because the two tier structures are not symmetrical. See what a ChatGPT subscription actually costs.
• If you stop subscribing, the money on the card comes back: the card balance can be pulled back to your account balance, free of charge, but only 30 minutes after the last transaction on that card; $1 stays on the card and comes back with the balance when you close the card. Returning your account balance to your own USDT wallet is called a prepayment refund, a flat $2 per transaction including payment processor fees — this is not a withdrawal, and your account is not closed.

9. FAQ

10. Related Reading

• ▸ How Claude usage limits work and what to do when you run out — the other half of the job, once the tier is settled: getting real efficiency out of the allowance
• ▸ Getting Claude Code running from China — if Claude Code is your main tool, read this before you settle on a tier
• ▸ Troubleshooting a declined ChatGPT subscription payment — the long version of the triage table, matched line by line against error messages; it is a ChatGPT article, but the troubleshooting logic applies to Claude just as well
• ▸ What a ChatGPT subscription actually costs — if you subscribe to both, put the two bills side by side first
• ▸ What is a virtual credit card — start here if this is the first you have heard of them