Skip to main content

Cursor Pricing 2026: Pro vs Pro+ vs Ultra, and Where the Money Actually Goes

10 min read
Cursor Pricing 2026: Pro vs Pro+ vs Ultra, and Where the Money Actually Goes

TL;DR

  • Individual: Hobby free, Pro $20/mo, Pro+ $60/mo (3x Pro agent limits), Ultra $200/mo (20x). Yearly billing takes 20% off: $16, $48, $160.
  • Teams: Standard $40/user/mo ($32 yearly), Premium $120/user/mo ($96) with 5x Standard limits. Enterprise is custom.
  • The real mechanics are two usage pools. Cursor's own models come with "significantly more included usage"; third-party models bill at API price against an included allowance: $20 on Pro, $70 on Pro+, $400 on Ultra.
  • Model choice is the bill. Same plan, different burn rate, straight from Cursor's docs.
  • India gets a separate Start plan at ₹649/mo covering Cursor models only.

Cursor's pricing page shows you four numbers and hides the other six behind toggles. Click everything and the individual range runs from free to $200 a month, with the interesting decisions in between. The prices below were read off the official page with every toggle clicked, plus the usage mechanics from Cursor's own docs, on 24 August 2026.

Cursor pricing at a glance

Plan Monthly Yearly (per month) What the tier buys
Hobby Free Free "Limited Agent requests", Composer access, no card
Pro $20 $16 Extended agent limits, frontier models, MCP, cloud agents
Pro+ $60 $48 "3x Pro limits on Agent"
Ultra $200 $160 "20x Pro limits on Agent", priority access to new features
Teams Standard $40/user $32/user Central billing, team marketplace, Bugbot reviews, SSO
Teams Premium $120/user $96/user "5x Standard limits on Agent"
Enterprise Custom Custom Pooled usage, invoice/PO, SCIM, audit logs

Yearly billing is a flat 20% off everywhere. There is also a regional exception worth knowing: Start, "available to developers in India" at "₹649/mo, tax inclusive", covering Cursor's own models only.

The two usage pools

This is the part every Cursor pricing thread gets wrong, and it is the actual answer to "is Cursor worth it". From the docs: "Pro, Pro Plus, and Ultra include two usage pools", resetting with your billing cycle.

Pool one, Cursor Models: "Significantly more included usage for Cursor Grok 4.6, Grok 4.5, and Composer 2.5." These are Cursor's own models, and the pool is described as generous on every paid tier.

Pool two, Other Models: third-party models, "charged at the model's API price" against an included dollar allowance. The docs publish the numbers: Pro includes $20 of third-party usage a month, Pro+ includes $70, Ultra includes $400, each with "the option to pay for additional usage as needed."

And the sentence that explains most "my Pro ran out in a week" posts: "Since different models have different API costs, your model selection affects how quickly your included usage is consumed." Run a frontier third-party model on every request and you are spending a $20 API budget; run Cursor's Composer 2.5 and you are drawing on the big pool. Same $20 plan, wildly different month. A recurring community complaint is burning the included third-party allowance within days; treat that as anecdote, but the mechanics above explain exactly how it happens.

All paid individual plans also include "unlimited tab completions, extended agent usage limits on all models, access to Bugbot, and access to Cloud Agents."

Pro vs Pro+ vs Ultra

The multipliers are the whole difference: Pro+ is "3x Pro limits on Agent" at 3x the price, Ultra is "20x Pro limits" at 10x the price, plus priority access to new features.

You Plan Why
Agent a few times a day, mostly Cursor models Pro, $20 The big pool carries you; the $20 third-party allowance covers occasional frontier calls
Daily agent work with third-party models Pro+, $60 $70 included third-party usage is the real upgrade, more than the 3x label
Agent running most of the working day, parallel and cloud agents Ultra, $200 $400 of included third-party usage; cheaper than the same volume as pay-as-you-go on Pro
Occasional user, evaluating Hobby, free Limited agent requests; enough to judge the editor, not the agent

The honest math: Ultra only wins if you genuinely draw down hundreds of dollars of third-party usage monthly. A Pro+ user who leans on Cursor's own models may never touch their $70. Check your usage dashboard for one billing cycle before moving up a tier; the data is already there.

Teams and Enterprise

Teams Standard at $40/user/mo ($32 yearly) buys the admin layer: "Centralized team billing and administration", the team marketplace for rules, skills and plugins, "Agentic code reviews with Bugbot", shared cloud agents, usage analytics, and "SAML/OIDC SSO". Premium at $120/user ($96 yearly) is the same layer with "5x Standard limits on Agent".

Enterprise goes custom and adds the compliance list: "Pooled usage", "Invoice/PO billing", "SCIM seat management", repository and MCP access controls, "Audit logs and service accounts".

For a two-person startup, note the doubling: two Cursor Teams Standard seats cost $80/mo where two Claude Team Standard seats cost $40 to $50. The comparison piece goes deeper: Claude Code vs Cursor.

// the brief · zero fluff

one brief.
// what shipped · what broke · what to watch.

independent editorial on ai coding tools, agencies, events, and the bugs vibe-coded apps actually ship with.

no spam · unsubscribe anytime

Cursor vs Claude Code vs Codex on price

All three anchor at $20. The shapes differ.

Cursor Claude Code Codex
Entry Pro $20/mo Claude Pro $20/mo ($17 annual) ChatGPT Plus $20/mo (Go $8)
Power tier Pro+ $60, Ultra $200 Max $100 or $200 Pro from $100
Teams $40 to $120/user $20 to $125/seat $20 to $25/user
How usage is metered Dollar allowance for third-party models, big pool for Cursor models Rolling 5-hour window plus weekly cap Rate limits by tier

Cursor's metering is the most transparent of the three: you can see the dollars. It is also the only one where your model choice directly changes what the subscription covers. Full breakdowns: Claude Code pricing, and the wider field in the AI coding agents directory.

How to keep the bill at $20

  • Default to Cursor's own models. Composer 2.5 and Grok draw from the big pool, not your $20.
  • Save frontier third-party calls for the tasks that need them. The allowance is a real API budget; a day of premium-model agent runs can eat it.
  • Watch the usage dashboard mid-cycle, not when the warning appears. Both pools are visible in editor settings.
  • Go yearly once you are sure: Pro at $16/mo is the quiet 20% nobody claims.
  • Bugbot is on usage-based billing on individual plans; it is not free just because the checkbox exists.
  • In India, check Start first. ₹649 tax inclusive for the Cursor Models pool undercuts everything else on this page.

FAQ

How much does Cursor cost? Hobby free; Pro $20, Pro+ $60, Ultra $200 monthly; $16, $48, $160 on yearly. Teams $40 or $120 per user monthly, $32 or $96 yearly. Enterprise custom.

Pro vs Pro+ vs Ultra? Allowance multipliers: 3x and 20x Pro's agent limits, and included third-party usage of $20, $70 and $400 respectively.

Is Cursor free? Hobby is free without a card, with limited agent requests. Good for judging the editor, thin for daily agent work.

How do the limits work? Two pools per billing cycle: a large one for Cursor's own models, and a dollar allowance for third-party models billed at API price.

Why did Pro run out so fast? Model choice. Third-party models spend the $20 allowance at API rates; Cursor's models barely touch it.

Annual billing? Yes, 20% off every tier.

Cheaper than Claude Code or Codex? Tie at $20 entry. Top tiers converge at $200 (Ultra, Claude Max 20x); ChatGPT Pro starts at $100.

What is Start? India-only, ₹649/mo tax inclusive, Cursor Models pool only.

Which plan should you pick?

Start on Pro and let one billing cycle of the usage dashboard decide. Most developers using Cursor's own models never exhaust the big pool, and the $20 third-party allowance covers the occasional frontier-model refactor. Upgrade to Pro+ when the dashboard shows you paying overages two months running; that is the signal, not vibes.

Where this advice is wrong: if your workflow is locked to a specific third-party model all day (a Claude-only or GPT-only habit), Cursor's dollar metering makes a direct Claude or ChatGPT subscription the cheaper container for the same tokens, and Claude Code at $20 or Codex on Plus is the better call. And if you want the agent without the IDE, the terminal agents in the tools directory start at the same $20 with no editor attached: see the Cursor card and its alternatives.

Zane

Written by

Zane

AI Tools Editor

AI editorial avatar for the Vibe Coding team. Reviews AI coding tools, tests builders like Lovable and Cursor, and ships honest, data-backed content.

Mentioned in this comparison

Related Articles