Cursor raises Teams limits and adds a Premium tier for power users

Cursor, maker of the popular AI coding environment, has announced higher usage limits for its Teams plan and a new Premium tier aimed at users who put the heaviest load on AI agents. The changes take effect immediately for all new customers. For existing customers, they apply from billing cycles starting July 1, 2026.

What changes in Teams plans

The main change is a separate usage pool for Composer, Cursor's latest model. According to the company, Composer 2.5 delivers frontier performance at a significantly lower cost than previous flagship models. Each Teams seat now gets two separate usage pools: one for Cursor's own models and Auto mode, and one for third-party APIs (Third-Party API), meaning models from other developers accessed through Cursor. The company says this increases the usage available to a standard seat without changing its price.

How the new Premium tier works

According to Cursor's recent Developer Habits Report, a small share of "power users" in any team accounts for most of the spending and the unpredictable costs beyond plan limits. To reduce those costs and make budgets more predictable for such users, the company is introducing the Premium seat: it includes 5x more usage than Standard but costs only 3x as much. Teams can freely mix seat types. Cursor estimates that the Composer pool in Premium should last a full month of intensive agent work for 99% of users.

Updated Teams plan terms:

• Standard: higher shared usage; $32 per seat per month with annual billing, $40 with monthly billing.
• Premium: 5x the usage of Standard; $96 per seat per month with annual billing, $120 with monthly billing.
• Changes apply to new customers immediately, and to existing customers from billing cycles starting July 1, 2026.
• The key model for the calculations is Composer 2.5.
• The source of the user-behavior data is the Cursor Developer Habits Report.

Why the new spending transparency matters

Beyond the plan changes themselves, Cursor has updated its dashboard, which now shows users how close they are to their limits, separately for the Auto + Composer pool and for third-party API models. The system will also recommend to administrators which seat type best fits a given employee based on their actual usage pattern. The company has also reworked its spending alerts: administrators can now set "smart" thresholds in dollars and receive Slack or email notifications before a billing-period invoice turns into a surprise.

Taken together, these changes reflect a broader problem facing every provider of agentic AI development tools: as agents take on more autonomous work, not just code autocompletion but full task execution, inference costs become far less predictable than under a classic subscription model. Separate pools, premium seats and proactive spending alerts are Cursor's attempt to combine flexibility for heavy users with budget control for teams and their finance departments.

Cursor's pricing changes also reflect the maturing of the AI development tools market. Early on, such products competed mainly on features and the quality of code autocompletion. By 2026, with agents taking on more independent work, from writing entire modules to multistep debugging, managing inference cost at the model level is becoming the key arena of competition. Companies building these coding environments must balance an attractive price for the mass user against covering the real compute costs of the small share of customers who push agents to their limits almost continuously.