Pricing guide · billed in dollars of AI usage
Cursor + Claude pricing & usage, explained (2026)
Short answer: Cursor Pro is $20/mo, and that $20 includes roughly $20 of model usage. Past the included allowance you keep going pay-as-you-go, billed at what the models themselves cost — no credits in the middle taking a cut (as of Aug 2026).
Feature for feature, the usage costs here are the lowest on this site: no platform mark-up sits between you and the model. Our receipts price Cursor at that usage rate rather than adding the subscription a second time — the $20 you pay each month is the first $20 of usage, not a fee sitting on top of it.
The other price is the learning curve. This path assumes you can read a diff (the list of lines the AI changed), run the app on your own machine, and tell a good change from a plausible one.
Cursor + Claude plans
| Plan | Price | What you get |
|---|---|---|
| Cursor Free | $0 | The editor plus a limited allowance of AI requests. Fine for evaluating, thin for building. |
| Cursor Pro | $20 per month | The subscription most people buy: the newest, most capable models inside the editor, with roughly $20 of model usage included in the price (as of Aug 2026). |
| Cursor Pro+ | $60 per month | Includes $70 of agent usage — billed at the models' own pay-as-you-go rates, what Cursor calls API usage — plus bonus usage, as published by Cursor (as of Aug 2026). |
| Cursor Ultra | $200 per month | Includes $400 of agent usage at the same pay-as-you-go rates, plus bonus usage, as published by Cursor (as of Aug 2026). |
| Usage beyond the plan | Pay as you go | Once the included allowance is gone, usage is billed afterwards at model rates — you pay for what the model reads and what it writes. |
| Claude Code | Subscription or pay-as-you-go | Anthropic's agent for the terminal — the text-only window where you type commands — billed either through a Claude subscription or pay-as-you-go by usage. |
Nothing to top up: usage is added to the bill as it happens. The discipline is yours to impose — set a spending limit on the account before a long session (as of Aug 2026).
How Cursor + Claude usage actually work
There are no credits here. You pay in tokens — the chunks of text the AI reads and writes: the code it reads out of your project plus the code it writes back. The $20/mo Pro plan front-loads about $20 of that, and anything past it is billed as usage.
That's why the per-feature numbers are the lowest on this site — you aren't paying a platform mark-up on top of the model, and you control the context: how much of your project the AI reads on each request.
One bill, not two. The receipts here count the $20 subscription as the first $20 of usage rather than adding it on top, which is why a short project can finish inside the plan and cost you exactly the subscription.
The cost that never appears on either invoice is your time. Budget for the hours as seriously as you budget for the tokens.
What features cost on Cursor + Claude
Base cost per feature in dollars of AI usage, before the iteration tax (the extra rounds nobody escapes) and the debug surcharge (fixing what breaks). These are the ranges our estimator adds up.
| Line item | Cost in dollars of AI usage |
|---|---|
| Core app — simple | $5–$11 |
| Core app — standard | $8–$18 |
| Core app — complex | $14–$28 |
| User accounts & login | $4–$9 |
| Payments / subscriptions | $6–$12 |
| Admin panel | $5–$10 |
| File & image uploads | $3–$6 |
| AI features | $6–$12 |
| Connections to other services | $5–$10 |
| Multi-user roles | $3–$7 |
Multiply by your prompting style (×1.2 to ×8), then add 40% for debugging. See the full formula →
The gotcha
What you spend depends on which AI model you pick. Cursor warns you near the included limit, and you can buy more or move up a plan. The learning curve is the real cost: you have to review the code and fix it when the agent goes wrong.
The cheapness is real but conditional. When the model gets it wrong here nothing catches it for you — there's no live preview window quietly confirming the app still runs, and a wrong turn costs you an evening rather than a credit.
Costs scale with tokens rather than messages, so a big codebase — all the files your app is made of — gets expensive to reason over even when your requests stay small. Keeping that context tight is the whole game.
The trade is consistent: a smaller usage bill, a longer first week. If you don't want to spend that week, one of the other five will be cheaper in the currency that actually matters to you.
Who Cursor + Claude is for
People who can already code, or are serious about learning. Small usage bill, steep curve, and a proper editor wrapped around the model instead of a bare terminal — which is the difference between this and Claude Code.
See real numbers
Printed receipts for three common builds on Cursor + Claude:
Get your build’s receipt
Pick what you’re building, pick how you prompt, and see the bill in Cursor + Claude usage before you spend any.
A plain link to the platform. Nothing on this page changes which one we call cheapest.