Last checked: 21 Oct 2026. Every fact below comes from Cursor's own documentation, help centre, pricing page, changelog and terms, and from the X Subscription documentation at docs.x.ai.
On Cursor's current plans the bill is built in a fixed order: a weekly pool of included usage for Cursor's own models, then prepaid credit that someone has to buy, then an optional monthly limit that caps paid usage. Which controls your team has depends on whether it is on that current plan family or a legacy plan, and several third-party explainers that rank for Cursor pricing searches still describe the older model. Cursor changes its prices often, so we describe how the charges work and link the pricing page for the figures.
First, work out which plan family you are on
New customers start on The X Subscription, the plan in which "Grok and Cursor share a weekly included usage pool." Our guide to SpaceX Super Intelligence (SpaceXSI) explains how Cursor, Grok and Grok Bot fit together.
Teams that bought earlier are probably on a legacy plan. Cursor's models and pricing docs say Hobby, Start, Pro, Pro+, Ultra, and Teams Standard and Premium seats "keep working as they are, with two monthly usage pools and on-demand usage." No retirement date is published. Switching is optional, and "The switch is permanent."
Cursor publishes no single check. Its help pages send current plans to grok.com/dashboard/billing and legacy plans to the Spending tab of the Cursor dashboard, so (our inference) the billing page you land on is a clue, and your Cursor admin can confirm it. Billing and some legal documents also sit on different sites while plans migrate, so confirm who your contract is with before a renewal.
How the bill is built on The X Subscription
| Layer | What it pays for | When it runs out |
|---|---|---|
| Weekly included pool | Cursor's own models, shared across Agent in the editor, the CLI, Cloud Agents, Grok Bot and the Grok apps. Unused usage does not roll over. | Work continues from prepaid credit, or you wait for the weekly reset, or you upgrade. |
| Prepaid credit (top-ups) | Usage beyond the pool, and third-party models at API list price on Premium and above. Charged when bought or when an automatic top-up runs. | With Auto Top-up off, additional usage stops. |
| Monthly additional usage limit | A cap on paid additional usage each month. Leave it unset for no cap. | Paid usage pauses until next month or until an admin raises it, even if credit remains. Included weekly usage keeps working. |
Sources: plan pricing docs, usage limits, overages, spend limits and Teams pricing, checked 21 Oct 2026.
The usage limits page lists Cursor's own models as Grok 4.7, Grok 4.6, Grok 4.5 and Composer 2.5 at the time of writing, and says "Third-party models aren't included." The entry Plus plan has no Auto, Cloud Agents, Automations, Bugbot or third-party models, and "Plus isn't available for teams."
The most useful point for predictability is Auto Top-up. The overages page says it is "off until you turn it on", that you are charged "when you buy credit or when an automatic top-up runs, not in arrears", and that additional usage "stops once your weekly pool and any credit you already bought are used up." Together, those mean that with Auto Top-up off, paid spend cannot exceed the credit an admin has already bought. Cursor also says "Requests are never downgraded in quality or speed."
Two details catch people out. Each top-up expires "currently 365 days after purchase". And the spend limits page warns that "Enforcement is not instant, so usage can briefly exceed your spend limit."
On a team
The Teams pricing docs say "Each paid seat has its own weekly pool for Cursor models" and "Seats don't share weekly pools with each other." Usage beyond a seat's pool is paid from a prepaid balance that "the whole team shares", so one heavy user's overflow draws on everyone's balance. That is why the team-wide limit matters.
On a legacy plan
The legacy plans page says on-demand usage on Teams is "On by default", that admins "can turn it off or set a team-wide spend limit", and that it is "billed monthly in arrears". On legacy Pro, Pro+ and Ultra it is off until you turn it on. A legacy team with on-demand left on and no limit set has an uncapped bill, paid in arrears. When a legacy team limit is reached, "all members consuming on-demand usage lose access to AI features."
Who can set limits, and at what level
On Teams, only admins change spend settings: "Only admins can buy credit or change Auto Top-up." They set one monthly additional usage limit for the whole team, "either a fixed amount or an amount per seat."
Whether Teams can also cap individual developers is where Cursor's pages disagree. The Teams pricing and Members pages describe a default per-member limit and per-member overrides. The spend limits help page describes member-level limits under Enterprise, and the Admin API marks its per-user limit calls "Enterprise only". Check which your plan shows in the admin dashboard before you rely on per-developer caps.
Enterprise adds pooled usage tracked "across the full contract term, not month by month", team-wide Dynamic Spend Limits "that automatically scale with team size", and both team-level and member-level limits. Cursor's rule for overlaps needs care: "When a user has limits from multiple sources, Cursor applies the highest applicable limit." An override can widen a cap as well as narrow it. Billing groups are for "reporting, internal chargebacks, and budgeting", and the page does not say they carry limits. Model controls that turn models on or off are Enterprise only.
Spend alerts count additional usage only, are available to Teams admins on The X Subscription, and "don't stop usage." The Grok Bot terms (updated 12 Oct 2026) also make the customer responsible for "configuring appropriate permissions, approval requirements, and spending limits", so the limit is your admin's job.
Model choice and Cursor Router
Cursor's available models page says "Your model choice directly affects how quickly your included usage budget is consumed."
Auto is now Cursor Router. The router docs say it "runs a classifier on each agent request" and "picks the most cost-effective model that still produces comparable quality for that request." That is Cursor's stated aim. The docs give no measured saving, and you influence routing by choosing a mode.
| Mode | Cursor's description | How it bills |
|---|---|---|
| Cost | "It optimizes token spend." | "bills as a Cursor-model request and draws from your weekly pool whatever model serves it." |
| Balance | "Optimizes for intelligence, speed, and cost." | At the routed model's list price. A route to a third-party model is paid from prepaid credit. |
| Intelligence | "Routes to the most capable models for harder tasks" | As Balance. Both "use your usage limits faster than Cost." |
Sources: router docs, router help page and models and pricing, checked 21 Oct 2026. The router page words billing more broadly ("All Auto modes bill at the list price of the model each request is routed to"), so we follow the more specific pricing page for Cost mode.
The router "is currently only available on Teams and Enterprise plans", and "Individual plans have Cost mode only." Enterprise teams must enable it because "it's off by default." At launch on 22 Jul 2026, Cursor's changelog said it was on by default for Teams. Balance is the default mode for new users.
On Teams and Enterprise, third-party requests also carry the Cursor Token Rate, a per-token charge "on top of the model's API price and paid from the team's prepaid credit." It applies whether a developer picks the model or Balance or Intelligence routes to it, and to your own API keys. Cursor's own models and Cost mode requests are exempt.
Team admins "can limit which modes members see and set the team default." An Impose Auto setting can default new chats to Auto or lock the picker to it, and both options are off by default. Cursor warns that "Blocking too many models reduces routing quality and can disable the router."
Matching the model to the job
Cursor's guidance on fit, as at 21 Oct 2026, is that Auto is "Good for everyday tasks", Grok 4.7 is "Built for the hardest, longest-running tasks", Composer is "built for interactive coding", and "Claude Opus and GPT-5.6 Sol handle complex, multi-step tasks well." That is the vendor's description. The right choice depends on your codebase and the work, and we do not rank models. Our post on reducing Claude API costs applies the same discipline to another vendor's billing.
Four levers sit beneath the model name:
- Third-party frontier models use credit faster. The Opus 5.5 page says it uses prepaid credit "faster than Sonnet 5.5 or Composer".
- Subagents can bypass the picker. They "can also run a named third-party model while the picker still shows Auto, Grok, or Composer", billed as third-party usage.
- Speed and effort. The Grok 4.7 page says "Fast is the default speed tier on Premium and higher", and Fast costs more than standard. Grok 4.7 also has four effort levels, with high as the default.
- Context. The cloud agent docs note that "a larger context window can increase token usage and costs."
Cloud agents, automations and Grok Bot
Cursor describes cloud agent billing in two ways. The plan pricing docs list Cloud Agents in the shared weekly pool. The cloud agent page says they are "charged at API pricing for the selected model" and that "You'll be asked to set a spend limit when you first start using them." Both may be true, but no page says so, so check how your usage page shows cloud agent runs before you budget.
The automations help says "Each automation run is a Cloud Agent run, billed at the API rate for the model selected." On a team, "Me bills to you, and Service account bills to the team", and only admins can choose the service account. A scheduled service-account automation is a standing team cost. Our post on cloud agents and automations covers what to automate first.
Grok Bot is included on Premium, Super and Ultra. Its usage "draws from the same weekly pool as Cursor and the Grok apps, not a separate Grok Bot meter", with extra usage paid from prepaid credit, and a limit is "not a hard stop in the middle of a run." Cursor's Grok Bot Teams docs say a separate Grok Bot spend cap is not available today, and that the dashboard usage page shows a split by product.
On Teams, Grok Bot "is enabled by default" and "There is no switch to turn it off". Enterprise admins can turn it off. Model choice is not a lever: the Grok Bot models page says "there is no model picker in Grok Bot settings" and that it "is not governed by your team's Cursor model allowlist or blocklist." Our comparison of Grok Bot and Cursor's cloud agents covers when each fits.
The Grok API is a separate bill that "doesn't draw from the pool." Our Grok API pricing post covers it.
A monthly review routine
This routine uses only views and settings Cursor documents, and covers spend and model mix. Whether the tool is improving delivery is a separate question, covered in measuring AI coding tools.
- Confirm the plan family and seat tiers. A seat on a lower tier may lack Cloud Agents, Grok Bot or third-party models.
- Pull spend from the Admin API. The Admin API returns spending for the current billing cycle: spendCents excludes included usage and overallSpendCents includes it. It is the documented source for per-member spend. The field descriptions use the legacy word "on-demand", which does not mean you are on a legacy plan.
- Check the prepaid balance, Auto Top-up and credit expiry.
- Check the team limit and each member's effective limit. On Enterprise, the highest applicable limit wins.
- Review the model mix. Look at the default router mode, any Impose Auto setting, third-party picks, subagents running named models, and the Fast speed tier.
- Check Grok Bot's share and every automation that bills the team.
- Check who receives spend alerts. They notify and stop nothing.
Name one owner for the routine, and one person who can approve raising the limit.
What Cursor's pages do not settle yet
As at 21 Oct 2026 we found no Cursor page that states a default additional usage limit for a new team on The X Subscription, gives a single account of how cloud agents are metered, draws a clear plan boundary for per-member limits, or says how Grok Bot's cloud computers count against a seat's allowance. The pricing page says "Prices subject to change", so re-read the linked pages before you set a budget. We plan to re-check this post by 21 Nov 2026.
Common questions
How does Cursor billing work for a team?
On The X Subscription, each paid seat has its own weekly pool for Cursor's own models. Usage beyond that pool, and third-party models on Premium and above, is paid from a prepaid balance the whole team shares, and only admins can buy credit or change Auto Top-up. Legacy Teams plans keep two monthly pools and on-demand usage billed in arrears, which is on by default.
Can I set a spending limit for each developer?
Cursor's Teams pricing and Members pages describe a default per-member limit and per-member overrides, while its spend limits help page places member-level limits under Enterprise and its Admin API restricts per-user limit calls to Enterprise. Check which your plan shows in the admin dashboard before you rely on it.
Does Cursor's Auto mode make coding cheaper?
Cursor Router has three modes. Cost mode bills as a Cursor-model request from the weekly pool. Balance and Intelligence bill at the routed model's list price and use limits faster. Cursor says the router aims for the most cost-effective model with comparable quality, but its docs publish no measured saving.
Do Grok Bot and cloud agents use the same allowance as the editor?
Grok Bot draws from the same weekly pool, with no separate meter and no separate spend cap today. Cloud agents are described two ways: Cursor's plan docs list them in the shared pool, and its cloud agent page says they are charged at API pricing for the selected model. Check how your usage page shows them.
What happens when we hit the limit?
Paid additional usage pauses until next month or until an admin raises the limit, even if credit remains, and included weekly usage keeps working. Enforcement is not instant, a Grok Bot run is not stopped mid-run, and spend alerts notify without stopping anything.
We use Cursor in our own engineering, and setting up plan family, limits, router defaults and automations is the kind of work our implementation service covers. If you want a view of your current setup, start with a diagnostic, or see our SpaceXSI and Cursor implementation service and Request a Quote.
John Kelleher holds the SpaceX Super Intelligence Partner Network Accreditation: Forward Deployed Engineer.
Get new articles by email
Practical guides from the SpotDev team, when we publish them.
By submitting this form, you consent to us sending you emails with our latest content. Privacy Policy
