In this guide
Cursor's pricing looks simple on the surface — Hobby, Pro, Pro+, Ultra — but the credit system underneath is what actually determines your bill, and it's the single most common source of confusion since the 2025 move to usage-based billing. This guide breaks down exactly what each plan includes, how credits are really consumed, and the habits that keep your monthly cost predictable.
The plans, in plain terms
| Plan | Price | What you get |
|---|---|---|
| Hobby | $0/mo | Limited Agent requests + Tab completions, no credit card required |
| Pro | $20/mo | Unlimited Tab completions, extended Agent limits, $20 monthly credit pool |
| Pro+ | $60/mo | Everything in Pro, roughly 3x the usage credits |
| Ultra | $200/mo | Around 20x Pro's usage, priority access to new features |
Pricing as of July 2026 (Teams at $40/user/mo also available). Verify current numbers on Cursor's pricing page.
How the credit system actually works
The most important thing to understand: Auto mode is unlimited on all paid plans and does not touch your credit pool. Credits are consumed specifically when you manually select a premium frontier model — Claude, GPT, Gemini — for a request. Since each paid plan's credit pool equals its price in dollars, your effective usage depends heavily on which models you choose to use deliberately.
This is why two Pro subscribers can have wildly different experiences: one who mostly uses Auto rarely touches their credits, while another who always manually selects Claude for every task can burn through $20 of credits quickly.
Tip
Claude models are typically the most credit-expensive per request, while Gemini tends to be the cheapest — a rough 2.4x difference is commonly reported between them. Reserve premium model selection for tasks that actually need the extra capability.
What changed in the 2025 billing switch
Before mid-2025, Cursor billed by a simpler request-count model. The move to credit-based billing tied cost more directly to actual compute (which model, how much context, how complex the task), which better reflects real usage but also made monthly cost less predictable for users who don't track which model they're selecting.
How to keep your Cursor bill predictable
- Default to Auto mode for routine work — it's unlimited on paid plans and doesn't touch your credit pool.
- Reserve manual frontier-model selection (especially Claude) for tasks that genuinely need the extra reasoning power.
- Scope agent tasks narrowly — see our Cursor Agent Mode workflow guide — since focused tasks use fewer tokens per request regardless of model.
- Monitor your credit usage in the first few weeks on Pro to calibrate what your specific workflow actually costs before assuming you need Pro+.
Which plan should you pick?
Hobby is enough to evaluate Cursor's agent quality on a small task. Most individual developers should start with Pro and rely on Auto mode by default — the $20 credit pool covers a meaningful amount of deliberate premium-model use for typical workloads. Only move to Pro+ or Ultra once you've confirmed from real usage that you're consistently hitting Pro's limits, not before.
Key takeaways
- Auto mode is unlimited on all paid Cursor plans and never touches your credit pool — use it by default.
- Credits are consumed only when you manually select a premium frontier model; Claude tends to cost more per request than Gemini.
- Each plan's credit pool matches its price in dollars ($20 Pro = $20 of premium-model usage).
- Start with Pro and rely on Auto mode; only upgrade to Pro+/Ultra once real usage proves you need it.
Frequently asked questions
Cursor has a free Hobby plan, Pro at $20/month, Pro+ at $60/month, and Ultra at $200/month, plus Teams at $40/user/month. Each paid plan includes a monthly credit pool equal to its price for premium model usage.