Cursor Pricing and Cursor Cost With Cursor AI Pricing for Pro, Teams and Enterprise

Cursor costs $20 per month for Pro, $60 for Pro Plus and $200 for Ultra on the individual side, and $40 per user per month for a Teams Standard seat or $120 for a Teams Premium seat. Enterprise has no list price. Every paid plan includes a monthly usage allowance, and anything beyond it bills as on-demand usage at the model's API rate, in arrears.

That second sentence is the part budgets miss. Cursor's own docs describe what real usage looks like on an individual plan: daily Tab users typically stay inside the included usage, daily agent users typically land at $60 to $100 a month in total usage, and power users running several agents or automations often pass $200 a month. So the plan names map to usage levels, not features. Pro Plus and Ultra buy a bigger allowance, and Cursor itself recommends Pro Plus for daily agent users and Ultra for power users.

Every plan now splits included usage into two pools, both reset each billing cycle:

When a Teams member uses up the Other Models pool, Cursor switches them to the Cursor Models pool. If they keep choosing a third-party model, or they exhaust both pools, they continue on on-demand usage, which is switched on by default for Teams. Quality and speed are never downgraded; the bill simply keeps running until a spend limit stops it.

Two ownership facts are worth a line in any vendor review. SpaceX completed its acquisition of Anysphere, the company behind Cursor, on August 14, 2026, and Cursor now runs as part of the SpaceXAI division. The Grok models in the first-party pool follow from that. Cursor also states it sells subscriptions only directly through cursor.com and authorizes no resellers, so a Cursor seat offered through a third party is not a legitimate discount.

Cursor Teams has two paid seat types. Standard is $40 per user per month ($32 annual) with the standard Teams allowance. Premium is $120 per user per month ($96 annual) with 5 times the Standard allowance. Teams can mix seat types freely, and a free Unpaid Admin seat lets a manager run billing and settings without holding a Cursor license. Both seats include centralized billing, usage analytics, team-wide privacy mode, SAML and OIDC single sign-on, Bugbot code reviews and cloud agents with shared team context.

The useful number is the unit price of usage. Premium buys 5 allowances for the price of 3, so each unit of included usage costs 40% less on a Premium seat than on a Standard one. That gives a clean upgrade rule you can apply from the spending tab: a Premium seat costs $80 a month more than a Standard seat ($64 more on annual billing), so once a developer's on-demand overage on a Standard seat passes $80 a month, Premium is the cheaper seat, as long as the four extra allowances cover that overage. Cursor says it expects the first-party pool on a Premium seat to cover a full month of heavy agent use for 99% of users.

Here is what the seat line costs a 100-developer engineering organization before any on-demand usage:

Annual billing takes 20% off both seats. Cursor bills per active paid seat rather than per pre-allocated seat, so adding a member mid-cycle creates a prorated charge. One rule is easy to miss: removing a member who has already used their allowance keeps that seat occupied until the cycle ends. Offboard contractors at the start of a cycle, not at the end.

What Teams does not give you is pooling. Each seat's allowance belongs to that user, does not transfer to a teammate, and resets every cycle. A team with a few heavy users and many light ones leaves unused allowance on the light seats while the heavy users run on-demand. That single fact is the main reason larger teams move to Enterprise.

Cursor Enterprise has no public price; you get a quote through Cursor sales. What Cursor does publish is how the contract is built, in its pricing policy, and that tells a buyer what to negotiate. An Enterprise order has two parts: a Core License fee per seat, and a Precommitted Usage amount, a minimum spend on model usage that is pooled across every user on the account. Usage above the commitment bills as on-demand at the rate in your order form.

The features Enterprise adds over Teams, per Cursor: pooled usage, invoice and purchase order 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, customer managed encryption keys and priority support. HIPAA business associate agreements are available to Enterprise customers.

The feature that changes cost control is spend limits. On Teams you can only set one team-level limit, and when it is reached, every member who is on on-demand usage loses AI features at once. On Enterprise you can set member-level and group-level limits, and on pooled accounts those limits apply to total usage, not just on-demand. For a US company with a few power users that is often worth more than the rest of the list.

The Cursor Token Rate is $0.25 per million tokens, charged on Teams and Enterprise for every request to a third-party model, on input, output and cached tokens alike, and on bring-your-own-key usage too. Cursor says it covers search infrastructure, custom model execution and routing. Cursor's first-party Grok and Composer models are exempt. Individual plans do not pay it.

Cursor passes third-party model prices through at list: the rates it shows for Claude Sonnet 5, Opus 5.5 and Fable 5.1 match Anthropic's own published API prices. So the token rate is the whole markup, and it behaves in a way a flat percentage would not. Because it is charged per token rather than per dollar, it weighs heaviest on cheap models and on cached tokens, and agent sessions are mostly cached tokens.

We priced one agent task on each model to show the effect. The task profile is our own illustrative assumption, not a Cursor figure: 300,000 fresh input tokens, 1.2 million cached input tokens and 40,000 output tokens, which is 1.54 million tokens in total, so the token rate adds $0.385. Cache write charges are left out. Rates are Cursor's published per-million-token prices.

Three findings hold whatever your exact token counts are. First, on a cached token of Claude Sonnet 5 or Opus 5.5, which lists at $0.20 per million, the $0.25 token rate more than doubles the price, which is why the markup shows up far more on agent sessions than on chat. Second, the cheapest third-party model is no longer cheap inside Cursor Teams: a GPT-5.6 Luna task that costs 13 cents at list costs 52 cents, about the same as Composer 2.5. Third, Grok 4.7 is exempt from the rate but not free of cost; its $0.50 cache read makes this task cost more on Grok than on Claude Sonnet 5 at list, though less than Sonnet once the token rate is added.

Two more line items belong in the model: opting in to regional data residency adds a 10% uplift to model pricing on eligible models, and Fast variants of the Cursor models cost more than their standard rates (Composer 2.5 Fast is $3 input and $15 output per million).

When you run out of included usage, Cursor keeps working and bills on-demand usage at the same API rates, monthly in arrears, as long as on-demand is enabled. On Teams it is enabled by default. Requests are never downgraded in quality or speed, so nothing in the editor tells a developer they are now spending real money.

Two details from Cursor's help center matter for finance. Enforcement is not instant, so usage can briefly run past the limit; Cursor credits that overshoot as a temporary spend-limit credit, but if you raise the limit in the same cycle it may bill some or all of it. And by default any team member can change the team's on-demand settings. An admin has to turn on the Only Admins Can Edit Usage Settings toggle under Permissions in the Spending tab to lock them down. Do that before the first invoice, not after it.

Most US teams pricing Cursor are comparing it with GitHub Copilot and Claude Code. The seat prices are close; the usage terms are not.

Read the table as a rule of thumb. On a Claude model, the same agent task costs the same list price on all three, plus 31% on Cursor Teams in our Sonnet 5 example. GitHub Copilot Business is the cheapest seat, and Cursor's first-party models are the cheapest way to run high-volume agent work inside Cursor. We ran the full Claude Code against Cursor comparison, including seat math for a 10-person team, in Claude Code vs Cursor pricing, and the Copilot side is broken down in GitHub Copilot Enterprise pricing.

There is one more comparison worth making before you size a Cursor contract. A lot of what an AI editor meters is repeated work that does not need a developer driving a frontier model at all: reviewing every pull request against your conventions, turning bug reports into labeled, routed tickets, and keeping API docs in step with the code. A ready-made agent that does one of those jobs runs on a flat monthly price with no tokens to watch. The coding agents at the top of this page, Code Reviewer, Bug Triager and Docs Writer, connect to GitHub and are priced per month, so they take the most predictable volume off your Cursor usage and leave the metered budget for real development.

If you are comparing coding tools across vendors, our Devin pricing breakdown covers Cognition's ACU model, Claude Enterprise pricing covers Claude Code seats and API-rate usage, and AI agent platforms compared puts the wider agent market on one price table.