Cursor Teams Pricing: Seats, Usage Pools & On-Demand Billing

Teams is self-serve: centralized billing, admin dashboard, SSOSingle Sign-On. One company login (usually via SAML or OIDC) instead of a separate password per tool. Press Enter for the full definition., analytics, team marketplace, BugbotCursor's automated PR reviewer that posts inline findings and can push fix commits from isolated VMs. Press Enter for the full definition. reviews and cloud agents. Enterprise (custom) adds priority support, invoicing, SCIMSystem for Cross-domain Identity Management. A standard for automatically creating and removing user accounts when people join or leave. Press Enter for the full definition., pooled usage, advanced security controls and per-member spend limits. Cursor recommends Teams for teams happy to self-serve and Enterprise when you need SCIM, pooled budgets or compliance features like CMEK and HIPAA BAA.

For finance the difference usually comes down to two items on that list: invoicing and pooled usage. Teams charges a card on a cycle, which works right up until a procurement department needs a purchase order and an invoice, and that requirement on its own is enough to move an org to Enterprise regardless of how the security list looks. Volume discounts on larger teams go through enterprise@cursor.com.

Neither plan makes you pre-buy seats.

Under roughly twenty seats, a card and a team-wide spend limit is the whole finance story, and Enterprise mostly buys paperwork you do not need yet. The threshold that decides it is not headcount. It is whether anybody at your company has to raise a PO to buy software.

Each paid seat comes with a monthly usage allowance split into two pools: one for ComposerCursor's own fast coding model, tuned for the editor and priced well below frontier models; the recommended day-to-day model for executing a plan. Press Enter for the full definition. and Auto, one for third-party API models. The allowance is per user by seat type, it does not pool across teammates and it resets every billing cycle.

• Included usage is split into ComposerCursor's own fast coding model, tuned for the editor and priced well below frontier models; the recommended day-to-day model for executing a plan. Press Enter for the full definition./Auto and third-party API model pools.
• Allocation is per user by seat type; it does not transfer between teammates and resets each billing cycle.
• Third-party model usage is charged at public list API prices plus the Cursor Token Rate when you exceed included amounts.
• The usage dashboard tracks both pools separately.

The two-pool split has one behaviour worth understanding before it confuses somebody on your team. Emptying the third-party pool stops nobody: they carry on against ComposerCursor's own fast coding model, tuned for the editor and priced well below frontier models; the recommended day-to-day model for executing a plan. Press Enter for the full definition. and Auto, so the models underneath their work change mid-month without anyone choosing that. An engineer reporting that the agent got noticeably different in week three is often describing a billing boundary rather than a model regression, and the pool split in the dashboard is where you confirm which one it was.

Since the allowance is per person and never transfers, the team average is not a useful number. The one to look at is the distribution, which is where the Premium candidates show up: the people who empty the third-party pool every month. Pooling one budget across everyone instead of splitting it per head is an Enterprise feature, covered in pooled usage.

That table is also the strongest argument for standardising on a model rather than letting everyone pin their favourite. At the time of writing it lists Composer 2.5The current Composer release, better at long-running tasks and at judging when a job needs a light touch versus deep work. Press Enter for the full definition. at $0.50 per million input tokens and $2.50 output, against $5 and $30 for GPT-5.5. An order of magnitude between two reasonable choices moves a monthly bill more than seat mix ever will. Cache reads are cheaper again, which is why switching models halfway through a thread costs more than it looks: the switch breaks the cache.

When a member exhausts third-party API included usage, Cursor can switch them to ComposerCursor's own fast coding model, tuned for the editor and priced well below frontier models; the recommended day-to-day model for executing a plan. Press Enter for the full definition. and Auto. If they keep using third-party models or burn through Composer/Auto included usage, on-demand usage continues service at the same rates, billed monthly in arrears, tracked per user in the admin dashboard. On-demand is on by default for Teams; admins can set team-wide spending limits.

That default is the setting to look at on day one if finance is nervous, because it decides whether the monthly bill has a ceiling before anyone chooses one. Leaving it on with a team-wide monthly limit keeps service continuous under a number you picked. Turning it off makes the included usage a hard stop, which arrives mid-sprint for whoever hits it first.

Hold two facts together, because they pair badly: on-demand bills in arrears, and per-member limits are Enterprise. On Teams you therefore learn the shape of a heavy month after it happened, at team level. The per-user spend tracking in the admin dashboard is what closes that gap, assuming somebody reads it weekly rather than at invoice time.

Non-Auto agent requests include a $0.25 per million tokens Cursor Token Rate covering semantic search, custom model execution (Tab, Apply) and infrastructure. It applies to input, output and cached tokens, including when you bring your own API keys. Budget models with this line item when comparing Cursor to raw API spend.

This earns its own row in a finance model for two reasons. Cached tokens count, so a long thread that mostly reads from cache still accrues the rate. And it applies to requests made with your own API key, which is the assumption that quietly breaks a build-versus-buy spreadsheet: the provider bill moves to your account and this line does not go away. The Cursor Token Rate page has the exemptions and the arithmetic.

You are charged for active paid seats, not a fixed seat count. Add, remove, upgrade or downgrade anytime; billing pro-rates within the cycle. Refunds appear as account credit on the next invoice. Your renewal date stays fixed.

That removes most of the forecasting problem and introduces a small accounting one. Refunds arrive as account credit rather than money returned, so a team that over-provisions in March and corrects in April carries the correction forward instead of watching it reverse. The renewal date holding steady through all of it is the detail that keeps a procurement calendar intact, and it is the reason mid-cycle changes are cheap to make.