Cursor Usage Limits: Weekly Pool, Top-Ups, Caps and Third-Party Models

Bottom line: how Cursor limits work now

Cursor has no fixed prompt count. Since October 12, 2026, new customers buy it through The X Subscription, and each paid seat gets one weekly usage pool shared between Cursor and Grok. The pool is stated only as a multiple of Plus, and different requests draw different amounts from it.

• One weekly pool, shared with Grok. Coding with Grok or Composer in Cursor, chatting with Grok, generating images or video and running Grok Bot all draw from the same allowance, which refreshes every week.
• Third-party models sit outside the pool. Claude, GPT and Gemini are billed at API list prices from the first request and paid from prepaid top-up credit. They are not available on Plus.
• When the pool runs out, further first-party usage is paid from top-up credit until the weekly refresh, so a heavy week costs extra rather than stopping work, unless your monthly spending cap or empty balance stops it.
• A spending cap is the control to set first. It limits paid extra usage in a calendar month and can stop requests even when prepaid credit remains.

Cursor limits by plan

SpaceXAI publishes no message or request counts for any tier. It says each new tier includes at least twice as much coding usage with its own models as the comparable old Cursor plan. Prices for annual billing are on Cursor pricing.

What draws from the pool

• First-party model requests. Agent and chat requests to Grok models and Composer draw from the weekly pool. Larger contexts, longer outputs and more tool calls draw more.
• Grok usage outside Cursor. Grok Chat, Imagine, Voice, Build and Grok Bot draw from the same pool, so a busy week in Grok leaves less for coding.
• Not the pool: third-party models. They are Additional Usage from the first request, paid from prepaid credit at the provider's API list price.

On a team, each seat has its own weekly pool, and seats do not share allowances. The workspace shares one prepaid balance for third-party models and extra usage, and admins manage seats, top-ups and spending caps from the dashboard.

Top-ups and spending caps

Extra usage on The X Subscription is prepaid rather than billed in arrears:

1. Prepaid balance. You add credit manually, or turn on Auto Top-up to buy credit automatically when the balance drops below a threshold you choose. Auto Top-up is off until you enable it.
2. Monthly spending cap. The cap counts paid extra usage, not the credit you add. Once it is reached, paid usage stops until the next month, even with credit left.
3. What stops a request. A first-party request stops when the weekly pool is used up and there is no credit or the cap is reached. A third-party request needs credit and room under the cap from the start.

For a developer who works mainly in Claude Sonnet 5.5 at $2 input and $10 output per million tokens, an Agent request that reads 40,000 tokens of context and writes 2,000 tokens costs about $0.10 in top-up credit, so a $30 monthly cap funds roughly 300 such requests on top of the subscription.

Older Cursor plans

Existing Cursor Pro, Pro+, Ultra and Teams subscriptions continue unchanged until you switch or cancel. They keep the earlier model: two monthly pools, one for Cursor models such as Composer and Grok and one for third-party models at API prices, a monthly reset with no rollover, and on-demand usage billed in arrears after the included amount, with a spend limit to cap it. Teams and Enterprise plans on the old model add a $0.25 per million-token Cursor Token Rate to requests made with your own API key.

If you switch to The X Subscription, your Cursor, Grok and X accounts are combined, the old plan is cancelled with a prorated refund and its included third-party usage ends. Developers who rely on Claude or GPT should estimate that usage at API list prices before switching, because it will no longer come out of an included pool. The switch cannot be undone. The X Subscription explained compares the old and new plans.

Practical strategies when the pool runs low

1. Use first-party models for routine work. Grok models and Composer draw from the pool you already pay for, while every third-party request costs top-up credit.
2. Keep heavy Grok use in check during coding weeks. Video generation and Grok Bot draw from the same pool as Cursor.
3. Set the spending cap before a long agent run and decide who on a team may raise it.
4. Move up a tier if you run out most weeks. Super has almost four times the Premium pool, and Ultra about three times Super.