Last updated: 1 October 2026
Quick Answer: Cursor pricing has two halves: a plan price for each user, and the model usage that plan includes. The plans differ mainly in the size of that allowance. Once a developer spends it, usage bills on demand at provider rates, and that second half decides a team's real bill.
The plan price is the start of the bill. Model usage beyond it is where team costs move.
Cursor is built by Anysphere and sold as a ladder of plans, each carrying a monthly allowance of model usage. For one developer the list price answers the question. For a team of fifty it does not: the second invoice line is usage.
How much does Cursor cost in 2026?
Cursor costs nothing on Hobby, a flat monthly price on individual plans and a per-user price on Teams. Enterprise is quoted, as Cursor's pricing page states.
On Cursor's models and pricing documentation, read on 1 October 2026, Pro costs $20 a month, Pro Plus $60 and Ultra $200. Teams Standard costs $40 a user a month and Teams Premium $120. The same page puts a daily agent user at $60 to $100 a month of model usage, and a power user at $200 or more.
| Plan | Who it fits | Included model usage | Spend controls |
|---|---|---|---|
| Hobby | Evaluating the editor | Limited agent requests, access to Composer | None needed |
| Pro | One developer, occasional agents | Both pools, standard allowance | Personal limit |
| Pro Plus and Ultra | Daily and power agent users | Both pools, larger allowances | Personal limit |
| Teams Standard | Most team members | An allowance per user, not transferable | Team-wide limit |
| Teams Premium | The heaviest users | Five times Standard allowance | Team-wide limit |
| Enterprise | Large organizations | Pooled across the team | Member limits, audit logs |
Plans as documented by Cursor on 1 October 2026; prices and links in the text.
Cursor models pricing and Cursor API pricing
Cursor models pricing splits into two pools that reset with the billing cycle. One covers Cursor's first-party models and carries the larger allowance. The other covers third-party frontier models at each model's published API rate, so an Opus-class model empties an allowance several times faster than a Composer-class one.
Cursor API pricing is not a separate tier. Third-party requests bill at the provider's list rate, and on Teams and Enterprise plans each also carries the Cursor Token Rate, a flat charge per million tokens. Your own provider key does not remove it.
What do Cursor's Hobby, Pro, Ultra, Teams and Enterprise plans include?
Hobby is the Cursor free plan, with limited agent requests. Pro, Pro Plus and Ultra are one product with progressively larger allowances. Teams adds administration; Enterprise adds pooled usage.
Cursor AI pricing by plan tier
Cursor Pro pricing buys the full agent experience with both usage pools; Cursor Pro Plus and Ultra buy the same with a larger allowance. For an individual, allowance size is the whole difference between Cursor pricing plans. Teams adds centralized billing, usage analytics, team-wide privacy enforcement and SSO. Enterprise adds pooled usage, SCIM provisioning, audit logs and model access controls.
Annual billing is the quiet discount. Cursor's June 2026 Teams pricing update set the Standard user at $32 a month on annual terms against $40 monthly, and Premium at $96 against $120, five times the usage at three times the cost. No Cursor Pro trial is published, so the cheapest evaluation is Hobby, then one month of a Cursor Pro subscription cancelled before renewal.
How does Cursor bill model usage beyond the plan?
Past the allowance, Cursor bills on demand at provider API rates with no markup, in arrears, on a separate invoice line. Teams have it on by default.
Usage does not roll over. Each cycle resets both pools, and an allowance belongs to the user it was issued to, so a quiet developer's leftovers cannot cover a heavy one.
Cursor Hobby plan and free plan limits
The Cursor Hobby plan is the free plan, and its limits are the point: limited agent requests and access to Composer. The Cursor AI Hobby plan suits a proof of concept; the Cursor Ultra plan sits at the other end, for agent power users.
Is the Cursor Pro plan unlimited?
No. The Cursor Pro plan includes unlimited tab completions, but model usage draws on a monthly allowance across both pools. When it runs out you enable on-demand billing, upgrade, or stop.
Cursor Pro limits and the Pro trial
Cursor Pro limits are measured in model usage, not requests, so two developers with the same Cursor Pro subscription hit different ceilings depending on the models they pick. With no Cursor Pro trial documented, treat the first paid month as the trial. So Cursor IDE pricing is largely a matter of model selection.
What does Cursor cost for a 50-developer team?
A 50-developer team pays a predictable seat line and a variable usage line. The seat line is arithmetic; the usage line is a pilot forecast.
Take fifty developers: forty on Standard, ten heavy agent users on Premium. On Cursor's team pricing documentation, read on 1 October 2026, that is $1,600 and $1,200 a month. The seat line is $2,800 before a token of on-demand usage. Annual terms cut it by 20%, and Cursor points larger teams to its enterprise contact for volume discounts.
| Cost line | What drives it | How to estimate yours |
|---|---|---|
| Standard seats | Headcount with Cursor access | Active developers, not licences bought |
| Premium seats | The few who run agents all day | Top fifth of the pilot's usage table |
| On-demand usage | Model choice and session length | Median usage times active days |
| Code review runs | Pull requests merged monthly | PR count times the per-review rate |
| Data residency | Regulatory scope, Enterprise only | Add the published uplift |
Cursor enterprise pricing
Cursor enterprise pricing is quoted, not listed, and the reason to pay it is usually governance. Enterprise pools usage across the team instead of fixing it per user, adds member-level spend limits and audit logs, and offers US-only data residency at a published 10% uplift.
What do Bugbot and other add-ons cost?
Bugbot, Cursor's pull-request reviewer, bills by usage rather than by user, on the same on-demand line. Budget it per pull request.
Cursor Bugbot pricing
Cursor Bugbot pricing changed in Cursor's May 2026 Bugbot update. That post retired the $40 per user a month subscription, moved reviews to usage-based billing, and puts the average review at $1.00 to $1.50 depending on the size of the pull request. For a team merging four hundred pull requests a month that is a real line that scales with output, not headcount.
What mistakes should you avoid when budgeting Cursor for a team?
The expensive mistakes hide usage until the invoice lands.
- Budgeting the Cursor team plan on seats alone. Seats are the floor; on-demand usage is the variable, and it is on by default.
- Buying Premium seats upfront. Pilot first, then upgrade only the handful who need five times the allowance.
- Leaving spend limits unset. Set the team limit at the forecast and restrict changes to admins.
If the constraint is where code runs rather than dollars, Cursor's documentation of 1 October 2026 describes Self-Hosted Machines, which move cloud agent tool execution to a machine you manage while Cursor still runs the agent loop and inference. Cursor alternatives for on-premise AI coding covers fully self-hosted options. The same arithmetic for Anthropic's assistant sits in our guide to Claude Code pricing, and the head-to-head sits on our Cursor vs Claude Code page.
How Origins AI prices the Coding Tool for engineering teams
Origins AI (originshq.com) builds the Origins AI Coding Tool, a self-hosted AI coding assistant and LLM gateway that runs on your own hardware, in your own AWS, Azure or GCP account, in hybrid form, or air-gapped with local models. An implementation team deploys it, so the cost takes a different shape: infrastructure and an engagement rather than a user line plus usage.
Origins AI does not publish a rate card. The company says the gateway tracks every model request by team, project and engineer and enforces token quotas and model restrictions, the cost-control job a hosted plan does with spend limits.
In on-premise and air-gapped modes, according to its product page, no source code is sent to any external service. In hybrid mode, the code context submitted to the model does leave your network. What an LLM gateway does for a budget is the same either way: one place for keys, quotas and logs.
For data residency, our comparison for enterprise teams with residency requirements works through that decision.
Hosted tools win on time to value. Self-hosting wins when usage is heavy and steady, when a security review blocks external inference, or when spend across several assistants needs one place to govern it.
Talk to an engineer
Send your team size, model mix and a month of usage data and we will cost it both ways. Book a call with an Origins AI engineer.