Cursor pricing: Plus, Premium, Super, Ultra and Teams, and what Claude costs inside

Cursor is now sold through The X Subscription, one subscription for Grok, Cursor and X. It has five individual plans — Free, Plus at $8 a month, Premium at $30, Super at $100 and Ultra at $200 — and Teams seats at the same prices. Yearly billing is $80, $300, $1,000 and $2,000 a year. This guide reads Cursor's pricing page, help and docs as they stood on 20 October 2026.

What the price includes

Every paid plan includes one weekly pool for Cursor's own models — Grok 4.7, 4.6 and 4.5 and Composer 2.5 — shared with Grok's Chat, Imagine, Voice and Bot. It refreshes weekly. Cursor states the plans only relative to Plus: Premium has four times its usage, Super fifteen times, Ultra fifty. It does not say how much that is in dollars.

Plus runs only those models, at fixed settings, with no cloud agents, Bugbot or third-party models. Premium and above add them. On those, Auto on an individual plan is Cost mode: it bills as a Cursor-model request and draws on the pool, whichever model serves it.

Claude, GPT and Gemini are billed on top

Third-party models are in no plan's pool. On Premium and above they bill at the model's API list price from the first request, paid from prepaid credit. Your own API key also works, for chat models but not Tab, and your provider bills it.

When the weekly pool runs out

You wait for the reset, keep going on prepaid credit — a top-up, or Auto Top-up when the balance drops below a threshold — or move to a bigger plan. Cursor models past the pool bill at their per-token rates. A monthly additional-usage limit caps what credit can be spent.

Cursor Pro, Pro+ and the legacy plans

Hobby, Pro at $20, Pro+ at $60, Ultra at $200, Start in India and the old Teams seats are now legacy plans. They are closed to new customers, and existing subscribers keep them until they switch or cancel; switching is optional and permanent. They keep two monthly pools and on-demand usage. Cursor names Premium as the comparable plan to Pro, and Super to Pro+.

Cursor Teams pricing

Teams is priced per seat: each member gets a Premium, Super or Ultra seat at $30, $100 or $200 a month, with the same weekly pool as the individual plan, and one team can mix them. Seats do not share pools; Enterprise, priced on request, pools usage. Teams created before The X Subscription keep their Standard ($40) and Premium ($120) seats.

On both business plans every third-party request carries a Cursor Token Rate of $0.25 per million tokens — on input, output and cached tokens, your own API key included. Grok, Composer and Cost mode are exempt. Admins set one monthly limit for the team's paid usage, plus per-member limits.

What Claude costs inside Cursor

Cursor lists Claude at Anthropic's list prices: Opus 5.5, Sonnet 5.5, Haiku 5.5 and Fable 5.1, and — hidden in the picker by default — Opus 5, Opus 4.8 and older models. It lists one cache-write rate, equal to Anthropic's five-minute write. On Teams the $0.25 rate is added to each, which more than doubles the cache-read price of Opus 5.5, from $0.20 to $0.45 per million tokens. That matters because an agent re-reads its context from the cache on every call.

What heavy use comes to can only be estimated here. Caprock does not read Cursor's agent; the estimate is my own Claude Code work on one machine at Anthropic's list prices: 56 active days from 18 July to 2 October 2026 came to $13,922, and the median day to $153. Most of those calls ran Opus 5 and 4.8, which Cursor lists at the same rates, but Cursor's agent builds its own context, so its calls will not match these one for one. It is not a Cursor bill. Every model's rate and a calculator: Claude API pricing.

Is Cursor Ultra worth it?

Ultra costs 25 times Plus and includes 50 times its usage; Super costs 12.5 times and includes 15 times, plus priority model access at peak times. Cursor recommends Super for daily agent users and Ultra for power users. Neither includes Claude: on Cursor it bills per token on top of the fee. At the same $200, Claude Max 20× is a flat fee for Claude only, with a five-hour window and a weekly limit. If your work is mostly Claude, the Claude plan covers it inside its limits; if it is mostly Grok and Composer, Cursor's pool does. Neither publishes its allowance in dollars, so the comparison stops there.

Claude Code inside Cursor

You can also run Claude Code itself in Cursor: its VS Code extension has an "Install for Cursor" link, and Anthropic says a Pro or Max plan "also covers Claude Code in supported IDEs, including VS Code, Cursor and other VS Code forks". That work goes to your Claude plan, not to Cursor's pools. Cursor vs Claude Code compares the two in full; Claude Code pricing covers the Claude plans.

See your Claude side

Caprock reads Claude Code's sessions from disk — in a terminal or inside Cursor, whether or not it started them — and prices every call per session, project and model, a list-price estimate as Claude Code's own /usage is, next to the plan's dollars and the share of the weekly limit the work used. It does not read Cursor's own agent; that stays on Cursor's usage dashboard. Free, open source; nothing leaves your machine: install Caprock.