Cursor usage limits, included usage and on-demand pricing on legacy Teams plans, explained

This page covers the structure Cursor now documents as legacy, for teams created before The X Subscription: Standard and Premium seats carrying two included-usage pools that reset with the billing cycle. “Teams created before The X Subscription have two paid seat types, Standard ($40/user/mo) and Premium ($120/user/mo, with 5x the Standard allowance)“¹. If your team is on the current plan, with weekly pools per seat and a prepaid balance shared by the team, that structure is explained on the current Cursor page.

On June 1, 2026 Cursor restructured Teams pricing into this shape. Each paid seat carries included usage in two pools, the Standard seat kept its price and gained more included usage, and a Premium seat arrived with five times the included usage of a Standard seat at three times the cost². A Standard seat is $40 per user per month, or $32 on an annual plan; a Premium seat is $120, or $96 annually¹². The change applied immediately to new customers and to renewing customers on billing cycles starting July 1, 2026².

Prices are exclusive of applicable taxes¹⁸, and subscriptions are sold only directly, through X Subscription Corp, with no authorized resellers¹⁸. Cursor never puts a number on the Standard allowance, only the word allowance and a multiple. What it does document is how the allowance is consumed, which is the part that decides your invoice.

Included usage and on-demand usage

Two units, and only one of them is money:

• Included usage is the allowance inside a seat. Each paid legacy seat carries it in two pools: Cursor Models, the first-party models, and Other Models, the third-party ones¹. Usage is metered in tokens at each model’s list price per million tokens, and “different models draw from the pool at different rates"³. The allowance belongs to the seat: usage “doesn’t transfer between team members"¹, it “resets monthly with your billing cycle"⁴, and “unused included usage does not roll over on either kind of plan"⁴. On Teams every member resets together on the team’s billing cycle⁴.
• On-demand usage is what a member spends to keep working after the included pools are consumed. It is billed monthly in arrears at public list API prices with no markup⁵¹³, plus the Cursor Token Rate of $0.25 per million tokens on “eligible third-party model requests”, which applies on Teams and Enterprise plans only and exempts the first-party models⁸. On-demand charges arrive as their own invoices and line items, distinct from the subscription⁵. On a legacy Teams plan, on-demand usage “is on by default, billed in arrears, and capped with a team-wide spend limit"¹; on legacy Individual plans it must be explicitly enabled in settings⁵. Team admins can disable it or set spend limits in the dashboard⁵.

On a legacy seat, when a member consumes all of their included third-party usage, Cursor switches them to the Cursor Models pool, and when that is gone too, they continue on on-demand usage if it is enabled¹. Requests are never downgraded in quality or speed³. The dollars start at the wall, and on Teams the wall is per seat.

Which plan you are on

The individual plans¹⁸ are not covered on this page, and neither is the current Teams structure: for the Premium, Super and Ultra seat tiers, weekly pools and prepaid credit, see the current Cursor page. Legacy request-based plans are being retired: a subscriber can keep one until the next renewal and is then migrated to usage-based pricing²³.

How the seat model changed in 2026

Under the per-seat model, one member’s heavy month is one member’s problem, and one member’s quiet month is nobody’s gain. The changes this year moved the walls and, on Enterprise, offered a way around them, without changing the shape on Teams.

• May 4, 2026. Enterprise admins gained model and provider allow lists and block lists, soft spend limits that warn instead of blocking, automatic alerts to users reaching 50%, 80% and 100% of a limit, and usage analytics filtered by user or product surface¹⁴.
• June 1, 2026. Teams pricing was restructured into two included-usage pools per seat, with a Standard seat getting more usage at the same price and a new Premium seat at “5x the included usage of the Standard seat, at only 3x the cost"². The stated reason: “a small number of power users on any team tend to drive the majority of spend and unpredictable on-demand costs"².
• July 22, 2026. Auto became Cursor Router, with Cost, Balance and Intelligence modes, available on Teams and Enterprise, on by default for Teams¹⁶. “Cost mode bills as a Cursor-model request and draws from the weekly pool, with no Cursor Token Rate”; “Balance and Intelligence bill at the list price of the model each request is routed to”, and a third-party route in those modes incurs the Token Rate¹⁵.
• August 12, 2026. Grok 4.6 arrived on every paid plan²² and sits in the first-party Cursor Models pool³.
• August 21, 2026. The Pricing Policy was last updated¹³. It is the contract behind everything above: model fees are total tokens times the model’s rate, model pricing is “subject to change in accordance with the Model provider’s list pricing”, and on-demand fees are billed monthly in arrears on actual consumption¹³.
• The X Subscription. Cursor now sells Teams on three paid seat tiers with weekly pools and a team-shared prepaid balance; teams created before it keep the structure this page describes¹. The vendor has published no date for the change, so this page does not give one. The current structure is explained on the current Cursor page.

Three details that matter more than the dates:

1. On Teams, the allowance belongs to the seat. Usage “doesn’t transfer between team members"¹. A member at the wall and a member with an untouched pool are the same team, the same month, and the same invoice.
2. A spend limit stops the person’s paid usage, not the bill. When a member reaches a spend limit, “paid additional usage stops for that specific user” until the limit is raised or the next period starts⁶. Enforcement is not instant, so usage can briefly exceed the limit; on legacy plans Cursor credits the pre-enforcement overage as a temporary spend-limit credit, and raising a limit mid-cycle can make that credited overage billable⁶.
3. The reset does not wait for anyone. The billing cycle starts on the date you subscribe and renews from there¹⁷; adding or removing members does not move it¹. Usage a member did not use by the reset is not theirs the following cycle, and on Teams it was never anyone else’s.

Why this cycle may look different

A team whose working pattern fit inside the old Teams allowance has a larger Standard pool on this structure, a Premium option, and a Router that is on by default²¹⁶. What the invoice does with that depends on who is on which seat and what they run.

Cursor’s June 1 announcement said the Premium seat covers “a full month of heavy agent usage for 99% of users"². The other 1%, and every Standard seat running agents, are the tail, and on a per-seat plan the tail is exactly the set of members who reach the wall.

What actually drains included usage

Model choice first. The docs publish a per-million-token price for every model, and a Fast variant costs more per token than the standard one: on Grok 4.6, twice as much³. “Cost mode bills as a Cursor-model request and draws from the weekly pool, with no Cursor Token Rate”; Balance and Intelligence bill at the list price of the model each request is routed to¹⁵. Grok 4.6 offers four effort levels, low through xhigh, so the same prompt can cost very different amounts of compute²². Total tokens include input, output, and cache reads and writes¹³. Two things that look like seat usage are not: Bugbot bills “from on-demand spend on a legacy Teams plan” rather than from the seat’s included usage¹⁹, and on Enterprise, service accounts consume the team’s usage pool without consuming a seat²⁰. Opting in to regional data residency adds a 10% uplift on model pricing for eligible models³.

“a small number of power users on any team tend to drive the majority of spend and unpredictable on-demand costs.” Cursor, announcing the June 1 Teams pricing²

What an administrator can do

Each control below does one thing and stops at a definite point.

• Roles. Members have no access to billing settings or the admin dashboard, and an Unpaid Admin has the same administrative capabilities without a billable seat or product access⁹. Whether Members can change on-demand and spend settings themselves depends on the “Only Admins Can Edit Usage Settings” toggle⁶. Where it stops: a role decides who may set a ceiling, not where usage goes.
• Team-wide spend limit. On a legacy Teams plan, on-demand usage is “capped with a team-wide spend limit"¹, set in the dashboard’s Spending tab and effective immediately⁶. Dynamic Spend Limits scale it with the seat count⁶. Where it stops: when the team limit is reached, every member consuming on-demand usage loses access to AI features⁶.
• Per-member and group limits. Member overrides, group overrides and the team’s general limit, with the highest applicable limit winning⁶, and soft limits that warn instead of blocking¹⁴. On pooled Enterprise accounts a member limit applies to total usage, not just on-demand⁶. Where it stops: a per-member limit decides when that member stops. It does not move what a quiet member left unused.
• Spend alerts. Email or Slack notifications when “additional usage spend” reaches a threshold, team-level or member-level, with the admin choosing who is told⁷². Where it stops: “Spend alerts don’t stop usage"⁷, and they count additional usage only, on-demand spend on legacy plans, never included usage⁷.
• The dashboard. On a legacy plan it shows “real-time usage for both pools, remaining allowance, and any on-demand charges"⁴, with on-demand tracked per user¹ and the billing view split into Included Usage and On-Demand Usage⁵. Where it stops: it shows. It does not decide.
• Analytics. A daily usage chart over the preceding 365 days, filters of up to ten users and 90 continuous days, CSV download per chart, and data collected only from clients on version 1.5 or higher¹⁰. Enterprise admins can filter by user or by product surface¹⁴. Where it stops: the same place.
• Model controls. Admins can force Router as a soft default or a hard lock²¹, and set allow lists and block lists at the model and provider level¹⁴. Where it stops: a cheaper model slows the meter for everyone equally. It does not change who runs out first.
• The Admin API. The page states no plan requirement for the API as a whole; the write endpoints are marked Enterprise only¹¹. It returns per-user on-demand spend in cents and total spend including included usage, daily per-user usage over at most 30 days, per-request usage events with model, tokens and charged cents, and it can set a per-user spend limit¹¹, at a default rate limit of 20 requests a minute and a recommended polling cadence of once an hour¹¹. Where it stops: it reads usage and writes ceilings. There is no call that moves allowance.
• Pooled usage. Enterprise’s answer to the per-seat wall: a shared pool instead of fixed per-user allocations¹². The stated reason is that without it “unused capacity from team members can’t be redirected to those doing the most demanding work"¹². The pool is a commitment for the contract period, tracked across the full term rather than month by month¹². Where it stops: precommitted usage not used within the term does not roll over¹³, and the same per-member limits still decide when a member stops⁶.

Five things to check before the next billing cycle

1. Confirm whether on-demand usage is on, and who can change it. On a legacy Teams plan it is on by default¹. Off means every seat stops at its second pool. On means the team-wide limit needs setting before the first member crosses, because a member past a limit is blocked until the next period⁶. Set the admin-only toggle if Members should not be the ones deciding⁶.
2. Re-forecast on the two-pool structure, not the old allowance. Standard seats gained usage, the Premium seat exists, and Router is on by default²¹⁶. Members who reached the wall under the old allowance may or may not reach it under this one; the dashboard shows both pools per member⁴.
3. Check the default routing and the model lists. Balance and Intelligence bill at the routed model’s list price while Cost mode draws from the pool as a Cursor-model request with no Token Rate¹⁵, a Fast variant costs more per token, twice as much on Grok 4.6³, and third-party tokens carry the Token Rate on top⁸. Cost mode and a block list are two of the few levers that work without changing anyone’s behavior.
4. Look at the distribution, not the average. On-demand is tracked per user in the admin dashboard¹, the analytics charts export to CSV¹⁰, and on Enterprise the Admin API returns each user’s spend in cents for the current cycle¹¹. A small number of power users drive the majority of spend²; the question is which of your seats they are, and which seats are sitting on an untouched pool.
5. If you are on Enterprise, know the term. The pool is committed for the contract period¹², unused precommitted usage does not roll over¹³, an invoice is issued for it at term end¹³, and if active users exceed the licensed seat count in a true-up period, extra seats and proportional usage are invoiced¹³. Renewal is automatic unless written notice is given at least 30 days before the expiry date¹³.

The part nobody has solved

Every native control, the on-demand switch, the team-wide, group and per-member spend limits, the alerts, the model lists, is a ceiling. Ceilings stop spending, and none of them moves unused included usage to the member who ran out. On Teams the shape is stated directly: usage “doesn’t transfer between team members"¹. Enterprise pooling removes the per-seat wall and replaces it with a contract-term commitment and the same per-member ceilings¹²⁶. Both results are true in the same organization, in the same cycle: allowance resetting unused on one seat, a blocked engineer on the next. The same is true of the other vendors' controls, which ship their own ceilings.

We are building Spillway Ops for that missing route: allowance that would otherwise reset unused flows to the members who ran out, so spend falls without anyone losing a tool. It reads usage metadata and sets per-user limits through vendor admin APIs with no repository access, no prompt contents, no code. The Cursor connection is planned. On Enterprise, the Admin API’s per-user spend reads and per-user spend limit writes are the documented path¹¹. If your team is weighing the move to The X Subscription, the current structure is explained on the current Cursor page, and we would genuinely like to hear how that conversation goes.