Cursor Pricing 2026: Every Plan, Usage Limits and Real Costs

Cursor pricing in 2026 starts at $0 for the free Hobby plan. Pro is $20/month, Pro+ is $60/month, Ultra is $200/month, Teams seats cost $40 (Standard) or $120 (Premium) per user/month, and Enterprise is custom-priced. Yearly billing takes 20% off every paid plan. The prices have not moved since 2025, but what they buy has: Cursor no longer gives each plan a fixed dollar credit pool, Auto is no longer free and unlimited, and since SpaceX bought Cursor in August, its own Grok models get the most generous limits.

We rechecked every number in this guide against Cursor’s pricing page and documentation on 2 October 2026. It walks through every Cursor AI plan, explains the two usage pools in plain English, shows what Bugbot and the new India-only Start plan cost, and compares Cursor with Claude Code and Devin Desktop, the tool formerly called Windsurf.

The Key Takeaways

• Free forever: the Hobby plan needs no credit card and includes limited Agent requests and access to Cursor’s Composer model.
• Pro is $20/month ($16 billed yearly) with unlimited Tab completions and two usage pools: generous limits on Cursor’s own Grok and Composer models, and a separate pool for Claude, GPT and Gemini at their API prices.
• Pro+ and Ultra are 3x and 20x Pro’s Agent limits. Cursor no longer publishes a dollar amount for any plan’s included usage.
• Auto is no longer free: every Auto request bills at the list price of the model it routes to.
• Bugbot is usage-based since May 2026, about $1.00 to $1.50 per review. The old $40 seat price is legacy.
• Student free year ended: Cursor closed its free-year-of-Pro offer to new signups on June 25, 2026; existing redeemers keep it until expiry.

Cursor AI Pricing Plans 2026 at a Glance

Cursor sells four individual plans worldwide, one extra plan for developers in India, and two business plans. Here is the full ladder, confirmed on the official Cursor pricing page and in Cursor’s models and pricing documentation.

Every price except Start is shown before tax. Cursor also sells only through cursor.com: its pricing page warns that subscriptions bought from resellers or third-party sellers are unauthorized and may be suspended at any time, so a cheap “Cursor Pro” from a key shop is a risk, not a bargain.

Since 26 August 2026, every paid individual plan and Teams carries one more perk: Cursor Pro now includes Grok Bot, the always-on agent app SpaceXAI and Cursor built together. Pro+ gets higher Grok Bot limits and Ultra the highest. That $20 entry matches Claude Pro, which is why we put Grok Bot up against Claude Cowork directly.

Cursor Hobby: the free plan (and what happened to the student year)

Yes, Cursor has a permanently free plan called Hobby. It includes a limited number of Agent requests every month and access to Composer, Cursor’s own coding model, and you do not need a credit card. Cursor lists access to frontier models from Anthropic, OpenAI and Google as a paid-plan feature. Hobby is enough to evaluate Cursor, but the limits make it unsuitable for full-time work.

For a year the far better deal was Cursor for Students, a full year of Pro for free worth roughly $240. That offer closed to new signups on June 25, 2026 after Cursor cited widespread verification fraud. Students who already redeemed it keep the free rate until their year expires, then roll over to standard $20/month Pro. Undergraduates now get credits and discounts at on-campus and online events this fall, and Master’s students, PhD candidates, researchers and educators can request credits through a form in Cursor’s help center. We break down exactly what is left in our Cursor student discount guide.

If you missed the free-year window, Hobby plus a rotation of competitor free tiers is now the cheapest route into a real AI coding workflow. It is also a useful way to learn how agent coding feels before paying for it out of your own pocket.

Cursor Pro at $20/month

Cursor Pro at $20/month ($16/month billed yearly, $192 a year) is the plan almost every paying user starts on. It includes unlimited Tab completions, extended Agent limits, generous limits on Cursor’s own Grok models, access to frontier models from Anthropic, OpenAI and Google, Grok Bot, MCP servers, skills and hooks, cloud agents, Auto, and Bugbot on usage-based billing.

In practice, how far Pro goes depends on which models you pick. Work on Grok 4.7 or Composer 2.5 draws from the larger Cursor Models pool. The moment you select Claude Opus 5.5, GPT-5.6 Sol or another third-party model, usage comes out of the separate Other Models pool at that model’s API price, and that pool runs out much faster.

If you run out before the month ends, you can turn on on-demand usage and keep going at the same API rates, billed monthly in arrears, instead of being forced onto a higher plan. Cursor says requests are never downgraded in quality or speed. That keeps occasional spikes cheap; if you blow past your included usage every month, the next tier up will probably save you money.

Cursor Pro+ at $60/month

Pro+ at $60/month ($48 billed yearly) adds no new features. It gives you 3x Pro’s limits on Agent and higher Grok Bot limits. Cursor itself recommends Pro+ for daily agent users, and its documentation puts a daily agent user’s typical total usage at $60 to $100 a month, which is why the plan sits at $60. For most students and hobbyists it is overkill.

Most readers will never need anything past Pro.

Cursor Ultra at $200/month

Ultra at $200/month ($160 billed yearly) is the heavy-hitter plan: 20x Pro’s limits on Agent, the highest Grok Bot limits, and priority access to new features. Cursor recommends it for agent power users, and its documentation says people running multiple agents or automation often use $200 or more a month. If your Pro+ bill regularly lands well above $150 with on-demand usage, Ultra becomes the more economical option.

For comparison, Anthropic’s Claude Max plan, which includes Claude Code, costs $100/month for the 5x tier and $200/month for the 20x tier. The 20x tier matches Cursor Ultra at the dollar but covers only Claude models. Many heavy users combine the two rather than picking one.

Cursor Start: the India-only plan at ₹649

On 28 July 2026 Cursor launched Cursor Start, a plan for developers in India at ₹649 a month, tax included, payable with UPI or a card. India is Cursor’s third-largest market, with more than 3 million developers. Start covers only the Cursor Models pool (Grok 4.7, 4.6 and 4.5 plus Composer 2.5) in non-fast mode at a fixed medium effort, along with cloud agents, Cursor for iOS, plugins, MCP servers, hooks and skills. It does not include third-party models, Auto, Bugbot, Automations, the Cursor SDK or on-demand usage.

Cursor Teams at $40 (Standard) or $120 (Premium) per user/month

Teams comes in two paid seat types. A Standard seat is $40 per user/month ($32 billed yearly), and a Premium seat is $120 per user/month ($96 billed yearly) with 5x the Standard Agent limits. Admins who only manage the team can take a free Unpaid Admin seat. Each paid seat’s usage belongs to that user, does not transfer between team members, and resets every billing cycle. Teams also pays a Cursor Token Rate of $0.25 per million tokens on every third-party model request, including bring-your-own-key use and Auto requests routed to a third-party model; Grok and Composer are exempt. When a member uses up their Other Models allowance, Cursor switches them to the Cursor Models pool, and on-demand usage is on by default. A 10-person team on Standard seats runs $400/month, or $3,840/year with yearly billing.

For small teams of 2 to 5 developers, individual Pro accounts plus shared documentation often work just as well. Teams earns its price once you need centralized billing, SAML or OIDC single sign-on, team-wide privacy mode, Bugbot code reviews on every pull request, or Cursor Router, the smarter version of Auto that is only available on Teams and Enterprise.

Cursor Enterprise pricing

Cursor publishes no Enterprise price; it is quoted by sales. The plan adds pooled usage shared across the whole company, invoice and PO billing, SCIM seat management, repository, model and MCP access controls, auto-run, browser and network controls, audit logs and service accounts, an AI code tracking API, controls for Grok Bot, per-member spend limits, and priority support with account management. The price depends on seat count, included usage and contract length, and larger teams can ask [email protected] about volume discounts.

Cursor Bugbot pricing

Bugbot, Cursor’s automated pull request reviewer, used to cost $40 per user per month. In its May 2026 update Cursor dropped the seat fee and moved Bugbot to usage-based billing; existing customers switched at their first renewal after 8 June 2026. Cursor says the average Bugbot run costs $1.00 to $1.50, depending on the size and complexity of the pull request. On individual plans it bills from your included usage, and on Teams it bills from on-demand spend. You can now set how hard Bugbot thinks per review, and there is a 14-day free trial. Its Autofix feature runs on cloud agents and bills at your plan’s rates.

How Cursor Usage Limits Work: Two Pools, Auto and the Router

This is the section every Cursor user wishes they had read first. Pro, Pro+ and Ultra each include two usage pools that reset with your monthly billing cycle. The Cursor Models pool covers Grok 4.7, Grok 4.6, Grok 4.5 and Composer 2.5 and comes with what Cursor calls “significantly more included usage”. The Other Models pool covers third-party models such as Claude, GPT, Gemini and Muse Spark, charged at each model’s API price. Both pools show up in your editor settings and on your usage dashboard. What Cursor no longer publishes is a size: until late August its documentation promised “at least $20” of third-party usage on every individual plan, and that line is gone.

Because each pool drains at the API price of the model you use, the model you pick matters more than the plan you are on. These are Cursor’s published rates for a selection of models.

Auto now has three modes: Cost, Balance and Intelligence. Whichever you choose, every Auto request bills at the list price of the model it is routed to, so Auto is a way to spend less per request, not a free lane. Balance and Intelligence run on Cursor Router, launched on 22 July 2026, which is only available on Teams and Enterprise and uses your limits faster than Cost. Max Mode, which used to expand the context window on any plan, now exists only on legacy request-based plans, billed at the model’s API rate plus 20%.

When you reach your included usage, you can either add on-demand usage at the same API rates or move to a higher plan. Cursor’s own rule of thumb: daily Tab users and light Agent users usually stay within their included usage, daily agent users typically land at $60 to $100 a month in total, and power users running several agents often pass $200. Your live numbers are on the usage dashboard, and limits reset at the start of each billing cycle.

What Changed in Cursor Pricing in 2026

Most of today’s pricing took shape between May and September 2026. The plan prices stayed the same; nearly everything around them moved.

• 11 May: Bugbot drops its $40 seat fee for usage-based billing.
• By late June: paid plans split into two usage pools, one for Cursor’s own models and one for third-party models at API prices.
• 25 June: the free student year closes to new signups.
• 22 and 28 July: Cursor Router launches with Auto Balance and Auto Intelligence, and the India-only Start plan arrives at ₹649.
• 14 August: SpaceX completes its acquisition of Cursor, which began as a SpaceXAI partnership in April. Grok becomes Cursor’s own model family.
• 26 August and 21 September: Grok Bot joins every paid plan, and Grok 4.7 becomes the newest model in the Cursor Models pool.

The June 2025 switch that started it

On June 16, 2025, Cursor moved from the old “500 fast requests per month” model to usage-based pricing, the model today’s pools grew out of. The reasoning, in Cursor’s own words on its official pricing announcement, was that newer AI models burn through tokens at vastly different rates depending on task complexity. The hardest requests cost the company an order of magnitude more than the simple ones, and the request-based model had stopped reflecting that reality.

The rollout was rough. Some Pro users hit unexpected charges in the first weeks because Cursor had not made it clear that “unlimited usage” only applied to Auto mode, not all models. Cursor publicly apologized and offered full refunds for surprise charges between June 16 and July 4, 2025 through [email protected].

If you joined Cursor in 2026, none of this affects you directly. It does explain the pattern: each change since has made Cursor’s own models cheaper to use inside the editor and third-party models a pay-as-you-go extra, a shift the SpaceX deal has only sharpened.

Cursor vs Claude Code vs Devin Desktop (Windsurf) Pricing

Cursor is no longer the only premium AI coding option. Here is how it stacks up against its two closest rivals at the individual and team level, with our take on Claude Code pricing as the reference point most Fello AI readers already know. Windsurf now belongs to Cognition, which renamed it Devin Desktop in June 2026. Google’s Antigravity is another option with a $0 plan, and its higher limits come with a Google AI Pro subscription.

Two patterns stand out. First, the entry plans are identical at $20/month and the top individual tiers at $200/month, so the choice is not about price at either end. Second, at team level Claude is now the cheapest per seat, because Anthropic includes Claude Code in every paid plan. A 10-developer team pays $250/month on Claude Team Standard, $400/month on Cursor Teams and $480/month on Devin Desktop. Only if every developer needs Claude’s 5x Premium seat does Claude jump to $1,250.

Seat price is not the whole bill. Cursor’s included usage is most generous on its own Grok and Composer models, while a Claude plan is all Claude, so a team that wants Claude Opus all day will usually get more of it per dollar from Anthropic directly. A team happy to work mostly on Grok and Composer, with Claude or GPT for the hard problems, gets more from Cursor. If you work on a Mac, our guide to OpenAI Codex on the Mac covers a third option.

Which Cursor Plan Should You Pick?

The honest answer changes a lot depending on who you are. We have grouped the four most common reader profiles below.

If you are a student. The free year of Pro is gone for new signups, so start on Hobby and lean on the free tier plus a few free trials of competitors. Our roundup of the best free AI for coding covers the strongest zero-cost options. If you are a grad student, researcher, or educator, check the credit request route in our Cursor student discount guide. Class projects rarely need more than Pro, and much of that work runs fine on Composer.

If you are a hobbyist coder or curious builder. Start on Hobby. If it gets in the way within the first week, jump to Pro. Pro is the right plan for evening tinkerers, weekend projects, and people learning to code with AI, and yearly billing brings it to $16/month.

If you are a side-project coder shipping real things. Pro is still your default, but watch the Other Models pool on your usage dashboard. If you reach for Claude or GPT on most tasks and keep paying on-demand on top of Pro, Pro+ at $60/month costs less once those extras pass about $40 a month. Skip Ultra unless you are running agents constantly.

If you are a full-time developer running agents all day. Pro+ is the floor, and it is the plan Cursor itself recommends for daily agent use. Ultra at $200/month makes sense once your total usage regularly passes $200 and you want predictable billing. Many full-time devs combine Cursor with Claude Code, using Cursor for fast utility work and Claude Code for deep multi-file refactors. See our roundup of the 30 best Claude Code skills if you decide to try the combo.

Tips to Stretch Your Cursor Plan

A few habits will keep your Cursor bill flat regardless of which plan you pick.

Default to Cursor’s own models. Grok and Composer draw from the larger Cursor Models pool, and Composer 2.5 costs $0.50 per million input tokens, a quarter of Grok 4.7’s rate. Use them for everyday edits, tests and small features.

Pick Claude or GPT on purpose. Every request to a third-party model comes out of the smaller Other Models pool at full API price. Save Claude Opus 5.5, GPT-5.6 Sol and Claude Fable 5.1 for the hard refactors and unfamiliar codebases where they actually change the result.

Watch the usage dashboard weekly. Cursor shows both pools live. Glancing at it once a week catches runaway agent sessions early and tells you whether you are trending toward a higher tier or fine where you are. If you turn on on-demand usage, set a spend limit with it.

Keep prompts tight and reuse outputs. Rambling prompts cost tokens. So does asking the same question twice because you forgot the answer. Save useful outputs into a personal notes file or build your own Cursor and Claude Code skills so the work compounds instead of evaporating.

Use yearly billing if you are committed. A 20% discount is one of the better deals in AI tooling right now. If you have used Cursor for three months and still love it, switching to yearly is almost always the right move.

Cursor for Non-Coders? Why Fello AI Might Be a Better Fit

Cursor is a code editor. If you do not write code, paying $20/month for an IDE you only use to chat with Claude or ChatGPT is the wrong shape of tool. The same is true for many of our readers who came to AI through writing, research, or productivity work rather than software.

If your goal is to chat with multiple AI models, generate text, summarize PDFs, write emails, build study materials, or get help with school and work, Fello AI is the better match. One subscription at $9.99/month or $79.99/year gives you Claude, ChatGPT, Gemini, Grok, DeepSeek and more in a single Mac and iOS app, with no terminal, no IDE and no usage pools to monitor. You pick whichever model fits the question, the way most people actually want to use AI.

For full-time coders, Cursor remains the right call. For everyone else, an editor-first tool is overkill. Compare what you actually want to do today, not what you might do in six months, and pick accordingly. Our best AI models in 2026 roundup is a useful next read if you are still deciding which model fits which task.

Conclusion

Cursor pricing in 2026 keeps the same price tags with new rules underneath. Hobby is free, Pro is $20, Pro+ is $60, Ultra is $200, and Teams runs $40 to $120 per seat. What you get for the money now depends on which models you use: Cursor’s own Grok and Composer models go a long way on any plan, while Claude and GPT draw from a smaller pool at API prices, and Auto no longer comes free. The student free year is gone. For most readers the right plan is Pro on yearly billing, or Hobby to test the waters first.

If you write code, start with the Cursor Hobby plan today and upgrade only if it gets in your way. If you do not write code, save yourself the IDE complexity and try Fello AI instead, where the same frontier models live inside an app built for chat, study, and productivity rather than software engineering.

FAQ