Cursor's paid plans give you two separate usage pools: a generous first-party pool for Cursor's own models (Auto, Composer 2.5, and Grok 4.5) and an API-model allowance for frontier models like Claude, GPT, and Gemini, billed at provider API rates.
| Plan | Price | API allowance | Key Features |
|---|---|---|---|
| Hobby | Free | — | Limited Agent + Tab completions, no credit card required |
| Pro | $20/mo | $20/mo included | Unlimited Tab, first-party pool, Bugbot access, Cloud Agents |
| Pro+ | $60/mo | $70/mo included | Everything in Pro with a larger allowance |
| Ultra | $200/mo | $400/mo included | Maximum allowance, priority access to new features |
| Teams | $40 (Standard) / $120 (Premium) per seat | Per-seat | Premium seats get 5x Agent limits; team marketplace, shared team context, analytics, team-wide privacy mode, SAML/OIDC SSO |
| Enterprise | Custom | Pooled usage | SCIM, audit logs, CMEK, repo/model/MCP access controls, AI code tracking API |
Official guidance from Cursor's pricing page: daily Tab users typically stay under $20/mo; daily Agent users typically land around $60–100/mo; power users running parallel and cloud agents budget $200+.
Max Mode unlocks a model's maximum context window and bills at API token rates. Cloud Agents always run in Max Mode — factor that in when sizing a plan for heavy cloud workflows.
Auto and Composer 2.5 for daily work — they draw from the generous first-party pool instead of your API allowance. Save Claude Opus-class models for the architecture decisions that genuinely need them, and use annual billing to save on the subscription.