Cursor Pricing: Every Plan, Limit and Price Explained

Cursor pricing changed in October 2026. New customers now sign up for "The X Subscription", which bundles Cursor with Grok and X Premium: Free ($0), Plus ($8 a month), Premium ($30), Super ($100) and Ultra ($200). The old Pro plan at $20 still works for people who already have it, but you can't buy it anymore.

The biggest change is how outside models are billed. Grok and Cursor's own Composer models come out of a weekly usage pool included in your plan. Claude, GPT and Gemini are not included at all: on Premium and up you pay their API list price from prepaid credit.

How we researched this: Cursor's pricing page, docs, help center and changelog as of October 2026, plus the public launch thread on Cursor's forum. We haven't run our own benchmarks.

Cursor pricing at a glance

These are the plans for new customers, from cursor.com/pricing and Cursor's pricing help page, checked October 24, 2026.

Prices exclude tax. Paying yearly saves two months on every paid tier: Premium is $300 a year instead of $360.

Because this is a bundle, the plans also include things that have nothing to do with coding. Plus lists X Premium perks like the checkmark and fewer timeline ads, along with Grok Chat, Voice and Imagine. If you only want a code editor, you're paying for those either way.

Is Cursor free?

Yes. Cursor's Free plan costs $0 and, according to its pricing help page, lets you use Agent, Chat and Tab completions with Cursor's own models (Grok and Composer). Cursor doesn't publish how many requests Free allows. It only calls the usage "limited".

Free does not include Claude, GPT or Gemini. It also skips cloud agents, automations and Bugbot code reviews.

Free is enough if you want to try the editor, use Tab autocomplete on side projects, or ask the agent a few questions a day. It stops being enough once you rely on the agent for real work every day, or you need a specific non-Grok model. For other zero-cost options, see our list of the best free AI for coding.

What each Cursor plan actually gets you

Plus ($8) is the cheapest paid tier, and it's more limited than the price suggests. Cursor's docs say Plus runs Cursor's models "at fixed settings (Fast off, medium effort)". It excludes Auto, Cloud Agents, Automations, Bugbot, Origin and third-party models. Think of it as a bigger Free plan for Grok and Composer.

Premium ($30) is the plan most developers should compare against other tools. It adds 4x the Plus usage, the latest Grok models, cloud agents and automations, code hosting and AI code reviews through Bugbot, and access to Claude, GPT and Gemini at API list price.

Super ($100) gives 15x Plus usage and priority model access at peak times.

Ultra ($200) gives 50x Plus usage and earliest access to new products and models.

On Teams, each seat costs the same as the matching individual plan and adds admin controls, SSO, usage analytics and central billing. Teams and Enterprise also pay a "Cursor Token Rate" of $0.25 per million tokens on top of the API price for eligible third-party models, per the models and pricing docs.

How Cursor's usage limits work

Every paid plan includes one weekly pool of usage for Cursor models. The usage limits help page lists those as Grok 4.7, Grok 4.6, Grok 4.5 and Composer 2.5. The pool is shared across Agent in the editor, the Cursor CLI, Cloud Agents, Grok Bot and the Grok apps.

Here's how it behaves:

• It refreshes weekly, and unused usage doesn't carry over.
• Different models draw from it at different rates. A Grok 4.7 request costs the pool more than a Composer 2.5 request.
• When it runs out, you can wait for the reset, buy prepaid credit at the same per-model rates, or upgrade. Prepaid credit lasts until used or 365 days after purchase.
• Cursor says requests are never slowed down or downgraded when you're near the limit.
• Max Mode is not part of the current plans.

Cursor doesn't publish the pool size in dollars or requests. So "4x Plus" tells you how plans compare to each other, but not how many agent runs you get. Your remaining usage and next reset date show in the editor settings.

One complaint from the launch thread on Cursor's forum is worth knowing: some users say the pool feels smaller because Grok Bot and the Grok apps draw from the same quota as Cursor.

What Claude and GPT cost inside Cursor

On the new plans, third-party models are pure pay-as-you-go. These are the per-million-token rates from Cursor's models and pricing docs, as of October 2026:

The Claude rates match Anthropic's own API list prices exactly. In our Claude Code pricing guide we worked out that one medium coding task (about 30 agent requests) costs roughly $0.99 on Sonnet 5.5 and $1.98 on Opus 5.5 at these rates. The same math applies inside Cursor.

So a developer on Premium who runs 100 of those tasks a month with Claude pays about:

• Sonnet 5.5: $30 plan + about $99 credit = about $129 a month.
• Opus 5.5: $30 plan + about $198 credit = about $228 a month.

That's the most important number in this guide. If most of your work runs on Claude, Cursor is now an editor on top of an API bill. A Claude subscription that includes Claude Code will usually cost less for the same model; our Cursor vs Claude Code comparison covers that trade-off.

How Cursor's usage-based pricing works

Cursor's usage-based pricing has two layers. Your plan's weekly pool covers Cursor's own models. Everything else, plus any Cursor-model use past the pool, comes from prepaid credit that you top up in dollars.

Per Cursor's plan docs, you control that spending in three ways:

• Manual top-ups, which last until used or 365 days.
• Auto Top-up, which refills your balance when it drops below a threshold you set.
• A monthly additional-usage limit, which works as a hard spend cap. On Teams, everyone shares one balance and one cap.

There's no markup on individual plans: third-party models cost their API list price. Two surcharges can apply, though. Regional data residency adds 10%, and some models charge double for very long prompts, such as Grok 4.7 above 256K tokens.

Is Auto mode unlimited?

No. On individual plans, Auto runs in a cost-saving mode and bills like a Cursor-model request, so it draws from your weekly pool. Plus doesn't include Auto at all. On older Cursor plans Auto was effectively unlimited, which is why many older guides still say so.

Does Cursor still use credits?

Not the way it used to. The current pricing page talks about a usage pool and top-ups. "Prepaid credit" in the docs just means the dollar balance you add. If a guide describes a monthly bundle of dollar credits, it's describing the plans from mid-2025, which only legacy subscribers still have.

Cursor CLI pricing

The Cursor CLI has no separate price. Its agent draws from the same weekly pool as Agent in the editor and cloud agents, per Cursor's plan docs. A long CLI session uses up your editor allowance for the week too.

Cursor doesn't say which plans can run the CLI. Plus is locked to Cursor's models at fixed settings, so expect Claude or GPT in the CLI to need Premium or higher.

Can you use your own API key in Cursor?

Yes, on most paid plans. Cursor's API keys docs support keys from OpenAI, Anthropic, Google, Azure OpenAI and AWS Bedrock, added under Settings, then Models.

The limits:

• Your key only works for chat and agent models. Tab autocomplete always runs on Cursor's side.
• Plus doesn't allow your own keys.
• On individual plans, requests on your key are billed by the provider and don't touch your pool or prepaid credit.
• On Teams and Enterprise, those requests still carry the $0.25 per million token Cursor Token Rate.
• Cursor's zero data retention promise doesn't cover requests sent with your own key.

So "Cursor API key cost" is whatever your provider charges. With an Anthropic key, Claude Sonnet 5.5 costs the same $2 input and $10 output per million tokens as it does through Cursor's prepaid credit. The real gain is using credit you already have with that provider.

What cloud agents and Bugbot cost

Cloud agents need Premium or higher. They bill at the chosen model's rate, draw from the weekly pool, and Cursor asks you to set a spend limit the first time you use one.

Bugbot, Cursor's AI code reviewer, moved to usage-based billing in May 2026 and is also Premium and up. Each review bills the account or team that turned Bugbot on for the repo. Cursor doesn't publish a per-review price, so watch your usage page for the first few weeks.

What happened to Cursor Pro at $20?

The Cursor Pro price was $20 a month, and existing subscribers keep it. Cursor's legacy plans page lists the old lineup: Hobby (free), Pro ($20), Pro+ ($60), Ultra ($200), Teams Standard ($40 per user) and Teams Premium ($120 per user), plus Start in India.

Legacy plans work differently. They have two monthly pools: one for Cursor models and one for "Other Models" like Claude and GPT, which is included on Pro, Pro+, Ultra and Teams. When you run out, you turn on on-demand usage.

New customers can't buy these plans. Switching to The X Subscription is optional, but Cursor's docs say the switch is permanent and support can't undo it.

Should you switch? Cursor says each new plan includes at least twice the coding usage with Cursor models compared with the legacy plan it replaces. That helps if you mostly use Grok or Composer. If you rely on Claude or GPT, the legacy plan's included "Other Models" pool is the reason to stay, since the new plans charge for those models from the first token.

When to upgrade (and when not to)

• Stay on Free if you're trying Cursor or coding a few hours a week with Tab and occasional agent help.
• Pick Plus ($8) only if Grok and Composer at medium effort are enough and you don't need cloud agents or code reviews.
• Pick Premium ($30) if Cursor is your main editor and you want the full agent features, or need Claude or GPT now and then.
• Move to Super ($100) when you hit the Premium weekly pool most weeks.
• Ultra ($200) is for people who live in the agent all day, every day.

Skip the upgrade if most of your agent work is on Claude. Every paid tier charges API list price for it, so a higher tier only buys more Grok and Composer usage. And if you want a different editor model altogether, our GitHub Copilot alternatives roundup lists other options.

FAQ

How much is Cursor per month?

As of October 2026, new Cursor plans cost $0 (Free), $8 (Plus), $30 (Premium), $100 (Super) or $200 (Ultra) a month. Yearly billing is $80, $300, $1,000 or $2,000.

Is Cursor free to use?

Yes, the Free plan includes Agent, Chat and Tab with Cursor's own models, with limited usage. Claude, GPT and Gemini need a Premium plan or higher.

What are Cursor's plans for teams?

Team seats cost $30, $100 or $200 a month, matching the Premium, Super and Ultra tiers. Enterprise pricing is custom and includes pooled usage, SCIM, audit logs and admin controls.

Does Cursor include Claude?

Not in the new plans' included usage. On Premium and up you can use Claude models at Anthropic's API list price, paid from prepaid credit. Legacy Pro, Pro+ and Ultra plans still include a monthly pool for third-party models.

Does the Cursor CLI cost extra?

No. The CLI uses the same weekly usage pool as the editor, so there's no separate CLI fee.

How do I set a spending limit in Cursor?

Set a monthly additional-usage limit in your billing settings. It caps how much prepaid credit Cursor can spend beyond your plan, including on Claude and GPT. Auto Top-up is optional and off unless you turn it on.