Cursor pricing: $8 to $200 a month, and Claude bills extra

How much does Cursor cost on each plan?

Cursor costs between $0 and $200 a month for one person, and teams pay $30, $100 or $200 a seat. Since 12 October 2026, new customers buy it as part of The X Subscription, which bundles Cursor with Grok and X Premium. We read every row below from Cursor's pricing page, its pricing help article and its models and pricing docs on 8 November 2026. US prices, before tax:

Premium is the plan Cursor marks "Most popular", and it's the cheapest one with Claude or GPT in the editor. Plus is labeled "Our introductory plan." It runs only Cursor's own models. Free gives you the editor with a small allowance and no third-party models. Cursor's pricing page lists no free trial for any paid tier. Yearly billing is ten months' price for twelve, so it saves $16 on Plus, $60 on Premium, $200 on Super and $400 on Ultra.

What does a typical month of Cursor cost?

A month of Cursor costs $30 on Premium if you stay on Cursor's own models, and $72 or more once you code with Claude. That's because the plan price buys only the weekly pool for Grok and Composer. Every Claude, GPT or Gemini token comes out of prepaid credit at the model's API list price. Here's one worked month at the rates Cursor published on 8 November 2026.

Say you send 20 agent requests a workday for 21 workdays. That's 420 requests. Each one reads 40,000 tokens of code and context and writes 2,000 tokens back. We counted every input token at the full rate, with no cache discount, so treat each figure as a ceiling:

The Sonnet row works like this. Claude Sonnet 5.5 costs $2 per million input tokens and $10 per million output in Cursor's price list. So 40,000 input tokens cost $0.08 and 2,000 output tokens cost $0.02. That's $0.10 a request, or $42 for 420, on top of the $30 plan. Opus 5.5 bills $4 and $20, which doubles it. Fable 5.1 bills $10 and $50.

Caching pulls the real bill down. Cursor charges $0.10 per million for cached Sonnet 5.5 input, a twentieth of the full rate. A long session that rereads the same files pays far less than the ceiling. The Composer row has its own catch. Cursor doesn't publish the size of Premium's pool in tokens or dollars, so we can't say whether 420 requests fit. If they don't, Composer 2.5 bills $0.50 in and $2.50 out per million, which would add at most $10.50 for this month.

Which Cursor plan is cheapest per unit of usage?

Ultra is the cheapest Cursor plan per unit of usage: $4.00 for each Plus-sized share, half of what Plus charges. Cursor sizes every tier as a multiple of Plus (4x, 15x and 50x), so dividing the price by the multiple gives a unit price Cursor doesn't print. We worked it out from Cursor's own list on 8 November 2026:

The jump that matters is Super to Ultra. Twice the price buys more than three times the usage. Below that, the tiers are close to flat: Premium saves you 50 cents a pool over Plus. So pick by how much you code, not by the discount. If Premium's pool runs dry two weeks a month, price the top-ups first. Super only pays off when you'd otherwise buy a lot of them.

What happened to Cursor Pro at $20?

Cursor Pro at $20 a month still works for people who have it, but Cursor no longer sells it to new customers. Its legacy plans page maps each old plan to the nearest new one, as of 8 November 2026:

For one person, Cursor's own models got cheaper and everything else got dearer. xAI's announcement says "Each X Subscription plan includes at least twice as much coding usage with our models as the comparable Cursor plan." Premium costs 50% more than Pro for at least twice the Grok and Composer usage. That puts a Pro-sized share at $15 or less. The catch is the second pool. Legacy Pro had one for third-party models, and Premium doesn't.

Teams came out ahead outright. A Standard seat at $40 became a Premium seat at $30, a 25% cut. Pro+ users lost their tier, since nothing sits between $30 and $100 now. If you're on a legacy plan, switching is optional and permanent. Cursor cancels the old plan with a prorated refund, and billing moves to grok.com. Cursor staff wrote in the forum announcement: "If you're on a current Cursor plan, your included third-party usage continues until you switch."

How much do Claude, GPT and Gemini cost inside Cursor?

Claude, GPT and Gemini cost their API list price inside Cursor, billed from the first request on Premium, Super and Ultra. Free and Plus can't use them. Cursor's models page prices them per million tokens. Here are the ones most people code with, read 8 November 2026:

Individual plans pay these rates and nothing more. Teams and Enterprise add a Cursor Token Rate of $0.25 per million tokens on eligible third-party requests. It skips Cursor's own models and any model whose cached input costs under $0.10 per million. On Auto, individual plans get only Cost mode, which Cursor bills as a Cursor-model request from the weekly pool, whichever model answers. Grok 4.7 doubles its rates past 256,000 input tokens.

What happens when you run out of Cursor usage?

When the weekly pool runs out, you buy a top-up or wait for the reset. Cursor's pricing FAQ says "Your allowance refreshes weekly, so one heavy day never wipes out your whole month." The pool is shared with Grok, so Grok Chat, Imagine, Voice and Bot all draw from the same weekly allowance as the editor.

Top-ups are prepaid credit. You can buy it by hand or turn on Auto Top-up, which refills when your balance drops below a line you set. A monthly additional usage limit caps the total. Cursor prints no top-up sizes or minimums on its pricing page or docs. Legacy plans work the other way round: two monthly pools, and on-demand usage billed after the fact under a spend limit.

What does Cursor cost next to the rest of your AI stack?

Cursor Premium plus ChatGPT Plus plus Claude Pro costs $70 a month, $840 a year at monthly prices. That's a common setup for developers who code in Cursor and chat elsewhere. We priced each plan on its vendor's page on 8 November 2026: ChatGPT Plus at $20 on OpenAI's pricing page, and Claude Pro at $20, or $17 a month billed yearly, on Anthropic's.

Look at the $70 row first. You're paying for three coding agents, because ChatGPT Plus includes Codex and Claude Pro includes Claude Code. If one agent is enough, drop two of them. Premium itself comes with Grok chat, images and video, so it may cover your chat too. And if you use Claude mostly inside the editor, a $20 Claude Pro plan doesn't pay for those tokens. Cursor bills them separately.

Whizi, which publishes this page, has no coding agent, no IDE and no API, so keep Cursor for the editor work. It's a chat app. Whizi Pro costs $29.99 a month for 2,500 credits and puts Claude, GPT and Gemini in one chat, with pictures included. It replaces the two $20 chat plans only if you don't need Codex or Claude Code. It can't pay for Claude inside Cursor, either. The cheapest way to use Claude and GPT runs the numbers for the chat side, and one subscription for every AI model shows what a single plan covers and what it doesn't.

Which Cursor plan should you buy?

Buy Premium at $30 if you want Claude or GPT in the editor, and Plus at $8 if Grok and Composer are enough. Start monthly. Move to yearly billing after two months, once you know your tier. The saving is two months' price.

Go to Super only if Grok and Composer top-ups on Premium pass $70 a month, the gap to $100. Claude and GPT bill the same on every tier. Go to Ultra when Super's pool runs out, since Ultra halves the unit price. Students lost their free year in June 2026, and Cursor's student discount covers what's left. Anyone still on legacy Pro at $20 should think twice before switching. Legacy plans are the last Cursor plans with Claude usage built in.