Cursor Pricing: Current Plans, Legacy Accounts and BYOK

Identify your account's plan first

Cursor's public pricing page now lists Free, Plus, Premium, Super and Ultra, with different model access and usage allowances. Its documentation also distinguishes legacy plans. An older article describing Pro and Pro+ is therefore not enough to identify what a new account buys.

Open your billing page and record the exact plan name, renewal price, billing interval and account type. Treat those account details as the starting point for a budget; promotions, taxes and negotiated terms can change the amount.

References: Cursor pricing

Separate included use from other model charges

The current pricing page describes a shared Cursor/Grok allowance and separately billed third-party model use on eligible plans. It also describes weekly refreshes for the current offering. Legacy accounts can have different mechanics. Do not apply an old monthly-pool explanation to every account.

Choose one representative task and model, note the usage before and after it, and inspect the resulting charge category. A pool multiplier cannot be converted into a reliable number of completed coding tasks without a measured workload.

References: Cursor pricing · Cursor models and pricing

Understand the BYOK exception

For eligible individual plans, Cursor documents provider-billed BYOK requests outside included usage. Teams and Enterprise have additional Cursor token-rate rules. Custom keys cover supported chat models rather than every feature. Requests still pass through Cursor's backend for final prompt building.

Check both the provider bill and Cursor's account settings. A working key does not establish that inference is unmetered, that your key funds completions, or that the request bypasses Cursor.

References: Cursor BYOK documentation

Make a budget from a week of real work

Record the fixed subscription fee, separate model charges and any provider-key costs. Include a short question, a multi-file edit and a longer investigation. Keep the same model and task boundaries when comparing alternatives.

Our suggested worksheet has five fields: task, model, accepted result, account charge category and cost. A failed task still belongs in the worksheet. Compare total spend with useful results, then choose a spend limit before expanding usage.