Cursor pricing in 2026 has five individual tiers and two business tiers. On Cursor's pricing page, checked on 24 October 2026, Free is $0, Plus is $8 a month, Premium is $30, Super is $100 and Ultra is $200. Teams seats are $30, $100 or $200 a seat a month, and Enterprise is priced on request. Every paid plan includes a weekly pool of usage for Cursor's own models, which Cursor shares with Grok because the subscription is now sold as part of one subscription with Grok and X. Third-party models such as Claude, GPT and Gemini are not in that pool: on Premium and above they are billed at API list price from prepaid credit, and Teams and Enterprise pay a further $0.25 per million tokens on most of them. This guide covers each plan, how usage is counted, and what a team of 5, 20 or 50 people would pay.
Every Cursor plan and price
Prices are in US dollars and exclude tax. Monthly prices are from the pricing page; annual prices are from Cursor's pricing documentation. Both were checked on 24 October 2026.
| Plan | Monthly | Annual | Included usage | Third-party models (Claude, GPT, Gemini) |
|---|---|---|---|---|
| Free | $0 | Not offered | Limited AI usage, agent and Chat requests | Not available |
| Plus | $8 | $80 a year | The base pool (1x), Cursor's own models only | Not available |
| Premium | $30 | $300 a year | 4x Plus | API list price, from prepaid credit |
| Super | $100 | $1,000 a year | 15x Plus | API list price, from prepaid credit |
| Ultra | $200 | $2,000 a year | 50x Plus | API list price, from prepaid credit |
| Teams | $30, $100 or $200 a seat | Not published | 4x to 50x Plus, pooled | API list price plus $0.25 per million tokens |
| Enterprise | Custom | Custom | Pooled across Cursor and Grok | API list price plus $0.25 per million tokens |
Paying annually brings each individual plan down by about a sixth: Plus works out at $6.67 a month, Premium $25, Super $83.33 and Ultra $166.67. Cursor does not publish an annual price for Teams seats.
What each plan adds
- Free gives limited AI usage and limited agent and Chat requests. It is enough to try the editor, not to work in it every day.
- Plus is Cursor's introductory plan. It includes only Cursor's own models (Grok and Composer) at fixed settings, so you cannot pick Claude or GPT on it.
- Premium adds 4x the Plus usage, the latest Grok models, cloud agents and automations, code hosting, AI code reviews and access to third-party models at API list price. Cursor recommends it for people new to AI coding.
- Super adds 15x the Plus usage and priority model access at busy times. Cursor pitches it at people who use agents daily.
- Ultra adds 50x the Plus usage and the earliest access to new products and models, for heavy users.
Because the subscription is sold through X Subscription Corp and bundled with Grok and X, the plans also list X features such as fewer timeline ads, creator tools and, from Premium, ad-free scrolling on most surfaces. If you only want the editor, those are along for the ride; there is no editor-only price. Cursor's FAQ says it has no authorised resellers, so buy only on cursor.com.
Is Cursor free?
There is a free plan, but it is limited. It includes limited AI usage and limited agent and Chat requests, and no third-party models. The cheapest paid plan is Plus at $8 a month or $80 a year, and the cheapest plan that lets you use Claude or GPT models is Premium at $30 a month or $300 a year.
How Cursor usage works
Cursor's pricing documentation describes two separate meters. Knowing which one a request lands on is most of the battle in predicting the bill.
1. The weekly pool, for Cursor's own models. Every paid plan includes one pool of usage for Cursor's models, which on 24 October 2026 were Grok 4.7, Grok 4.6, Grok 4.5 and Composer 2.5. The pool is shared across Agent in the editor, the CLI, Cloud Agents, Grok Bot and the Grok apps, so time spent in Grok chat comes out of your coding allowance. It refreshes every week, and different models draw from it at different rates. Cursor publishes the pool as a multiple of Plus (4x, 15x, 50x) and not as a number of tokens or requests, so no one outside Cursor can tell you exactly how far it goes. Your remaining usage and reset date are shown in the editor settings.
2. Prepaid credit, for everything else. Third-party models are never in the pool. From the first request, they are billed at the model provider's API list price and paid from prepaid credit. The same credit pays for Cursor's own models once the weekly pool runs out. You can turn on Auto Top-up to refill credit when the balance falls below a threshold, and set a monthly additional usage limit that caps spend regardless of how much credit you hold. On Teams, admins manage one shared balance and one cap.
When the pool runs out, you can wait for the weekly reset, spend credit or upgrade. Cursor's documentation says requests are not downgraded in quality or speed when that happens.
What third-party models cost in Cursor
Cursor charges the provider's list price, so the price of a Claude or GPT request in Cursor is the same per token as calling that provider's API directly. The examples Cursor's documentation gives are Claude 4.6 Sonnet at $3 per million input tokens and $15 per million output tokens and GPT-5.5 at $5 and $30. Check the provider's own pricing page for the model you plan to use, since the list changes often.
The Cursor Token Rate on Teams and Enterprise
Teams and Enterprise pay a Cursor Token Rate of $0.25 per million tokens on top of the list price for eligible third-party models. It applies to input, output and cached tokens, including when Cursor's router picks a third-party model for you. It does not apply to Cursor's own models or to individual plans, and Cursor excludes a list of small models from it, such as Claude Haiku 5.5, GPT-5 Mini and several Gemini Flash models.
The cached-token part matters more than it sounds. An agent session resends a large, mostly cached context with every request, so cached tokens can outnumber output tokens by a wide margin. As an illustration (our arithmetic), a developer whose agent sessions move 200 million tokens a month through Claude would add $50 a month in token rate, whatever share of those tokens were cache reads.
Legacy plans
Pro, Pro+, the old Ultra, Hobby, Start and the Teams Standard and Premium seats are legacy plans. Cursor says they continue for existing customers and keep working as they are, with two monthly pools and on-demand usage instead of top-ups. Max Mode is available only on legacy request-based plans and is billed at the model's API rate plus 20%. New customers cannot choose them.
What a team of 5, 20 or 50 pays
These are our arithmetic on the seat prices above, checked on 24 October 2026, billed monthly because Cursor publishes no annual Teams price. They cover seats only. Add third-party model usage at list price plus the token rate if your team uses Claude, GPT or Gemini.
| Team | All Premium-tier seats ($30) | Mixed seats | All Ultra-tier seats ($200) |
|---|---|---|---|
| 5 people | $150 a month, $1,800 a year | 4 at $30, 1 at $100: $220 a month, $2,640 a year | $1,000 a month, $12,000 a year |
| 20 people | $600 a month, $7,200 a year | 15 at $30, 4 at $100, 1 at $200: $1,050 a month, $12,600 a year | $4,000 a month, $48,000 a year |
| 50 people | $1,500 a month, $18,000 a year | 35 at $30, 12 at $100, 3 at $200: $2,850 a month, $34,200 a year | $10,000 a month, $120,000 a year |
Three things to plan for on top of seats:
- Model choice drives the real bill. A team that stays on Cursor's own models pays mostly for seats. A team that defaults to Claude or GPT for agent work pays list price for every request, so its usage line can match or pass its seat line. Set the monthly additional usage limit before the first invoice.
- Pooled usage evens out. Teams usage is pooled, so a few heavy users draw on headroom from light ones. That makes mixed seats work: give the $100 or $200 seat only to people who run agents all day.
- Enterprise is about controls. It adds invoice and PO billing, SCIM, granular spend limits by group, repository, model and MCP access controls, audit logs and service accounts. You need it when security or procurement requires those, not because of headcount.
Cursor vs Claude Code on price
The entry prices sit close together. Claude Code starts on Anthropic's Pro plan at $17 a month billed annually ($20 monthly), and Team seats start at $20 a seat billed annually, checked on claude.com/pricing on 24 October 2026. Cursor's cheapest plan that includes Claude models is Premium at $30 a month, and those models are billed on top at list price. The difference in practice is where model usage is paid: Claude subscriptions include Claude usage within their limits, while Cursor includes its own models and charges for everyone else's.
Plans, limits and API costs for Claude Code are covered in full in our Claude Code pricing guide, and the two tools are compared side by side on features, limits and fit in Claude Code vs Cursor. For a running list of AI coding tool prices, see the platform price tracker.
Which plan to choose
- Trying it out: Free, then Plus at $8 if you are happy with Cursor's own models.
- One developer who wants Claude or GPT: Premium at $30, with a low monthly additional usage limit while you learn what your sessions cost.
- Daily agent use: Super at $100 if Premium's pool runs out before the weekly reset most weeks.
- A team: Teams with mostly $30 seats and $100 or $200 seats for the heaviest users, a shared credit cap, and a check of the usage dashboard after the first month.
- Regulated or large organisations: Enterprise, for SCIM, audit logs and model and MCP access controls.
Taking a Cursor-built app to production
Cursor makes a working app cheap to build. What it does not do is decide whether that app is safe to put in front of real users and their data. That takes access rules that hold, secrets kept out of the client, tests, monitoring and a deploy you can roll back. Our Prototype Production Review checks an app built with Cursor, Claude Code, Lovable, Bolt or Replit for exactly that (security including Supabase row-level security, exposed keys and auth; the data model; performance; and deploy) for $2,500 fixed over 5 business days, and ends in a graded report and a fix plan. The fee is credited in full to a build with us started within 60 days. If you want engineers who already work this way to fix and ship it, our embedded teams cost $3,000–$7,500 per engineer per month. Many apps never need either: if it is an internal tool or a prototype, the AI-built app security checklist may be all you need.
