Cursor Pricing 2026: Every Plan Compared, and Where Startups Actually Save

Every Cursor plan in 2026, what "agent limits" actually mean for your bill, and why the real startup savings are on the model side rather than the editor.

Hobby $0 No card, limited agent requests
Pro $20/mo Frontier models, cloud agents
Teams $40/user Standard seat, per month
Startups Models No Cursor program; credits cover Claude and OpenAI

Cursor plans and prices (September 2026)

Prices are what Cursor lists on cursor.com/pricing for US customers before tax. Cursor changed its pricing model in 2025 from a fixed number of "fast requests" to usage-based agent limits, so older guides quoting "500 requests" are out of date.

PlanPriceAgent limitsWhat you get
Hobby $0 Limited No credit card. Limited agent requests, access to Composer, the full editor. Enough to evaluate Cursor, not enough to work in it all day.
Pro $20 / month Extended Frontier model access, generous Grok limits, Grok Bot, MCPs, skills and hooks, cloud agents, Bugbot (usage-based). The plan most individual developers are on.
Pro+ See pricing page (about $60 / month) 3x Pro Everything in Pro with three times the agent limits and higher Grok Bot usage. For developers who hit Pro's ceiling most days.
Ultra See pricing page (about $200 / month) 20x Pro Everything in Pro+ with twenty times Pro's limits, the highest Grok Bot usage, and priority access to new features. For all-day agentic work.
Teams Standard $40 / user / month Standard Centralized billing and admin, team marketplace, shared-context cloud agents, Grok Bot, Bugbot, analytics, team privacy mode, SAML/OIDC SSO.
Teams Premium See pricing page 5x Standard All Standard features with five times the agent limits and higher Grok Bot usage per seat.
Enterprise Custom Pooled Pooled usage across the org, invoice and PO billing, SCIM, access controls, audit logs, priority support. Contact sales.

Pro+, Ultra and Teams Premium prices are shown as descriptors ("3x", "20x", "5x") on Cursor's pricing page and the exact monthly figures change; the dollar amounts above are our latest check. Confirm on the pricing page before you subscribe.

What "agent limits" actually mean

This is the part that confuses people, because Cursor does not sell you a number of requests any more. Each plan includes a pool of usage measured against the cost of the models you call. A quick tab-completion costs almost nothing. A long agent run that reads twenty files, calls Claude Opus, and iterates on a failing test can cost a meaningful fraction of a day's allowance.

Three things move the bill more than anything else:

  • Which model you pick. Cursor routes to frontier models from Anthropic, OpenAI, Google and xAI. Opus-class models cost several times what Sonnet-class models cost per token, and that ratio is passed through to your limits. Auto mode picks a cheaper model when it can.
  • How much context you feed it. Agent mode with a large codebase and long conversation history sends far more tokens per turn than a single-file edit. Start new chats between unrelated tasks.
  • Background agents and Bugbot. Cloud agents and Bugbot run on top of your plan and Bugbot is billed on usage. Both are easy to leave running.

If you routinely see "you've hit your limit" mid-afternoon on Pro, that is the signal to move to Pro+, not to buy more Pro seats. If you hit it before lunch, Ultra.

Which Cursor plan should you pick?

Hobby if…

You want to try Cursor for a week. It is a full editor with limited agent usage, which is enough to decide whether the workflow fits. It is not enough for daily work.

Pro if…

You write code most days and use the agent for real tasks a few times a day. At $20 it is the same price as Claude Pro and cheaper than paying per token for the same models. Most solo developers land here.

Pro+ or Ultra if…

Agentic coding is your primary way of working. Cursor's Pro+ tier is the middle option for developers who exhaust Pro daily; Ultra is for people running background agents all day. Compare against Claude Code on Max before committing: for pure terminal-driven agent work Max 20x at $200 is the same price as Ultra and includes the Claude model usage directly.

Teams if…

You have more than a couple of engineers and want one invoice, SSO, privacy mode enforced for everyone, and analytics on who is using what. Standard seats are $40, double the individual Pro price, and that premium is for the admin layer.

Enterprise if…

You need pooled usage, SCIM, audit logs, or procurement wants a PO. Everyone else should stay on Teams.

For founders

Where startups actually save on Cursor

Cursor does not run a startup credit program and there is no Cursor listing in our directory that gives you free seats. Cursor has occasionally offered promotions to accelerator or VC portfolio companies, but nothing standing that you can apply for. Treat the seat price as a fixed cost.

The fine print

The cost that does scale is the models Cursor calls on your behalf. Cursor lets you bring your own API keys for Anthropic and OpenAI, and when you do, model usage is billed to those accounts instead of counting against your Cursor limits. That is where startup credits land:

Two caveats. Some Cursor features (background agents, certain model routes) require Cursor's own model access and will not run on your key. And bring-your-own-key means you are paying per token at API rates once the credits run out, so keep an eye on the usage dashboard.

Claude for Startups

Up to $100,000

  • $25,000 to $100,000+ in Claude API credits (partner tier)
  • Priority rate limits and early access to new models
  • Access to Claude Code, Claude API, Managed Agents and Cowork

Step by step: cutting your Cursor bill as a startup

Do these in order. The first three are free and take under an hour.

  1. 1

    Apply for Claude and OpenAI startup credits

    Both programs are self-serve. Claude for Startups takes about two minutes at claude.com/programs/startups if you were founded within the last four years. OpenAI for Startups is at openai.com/startups. Our Anthropic and OpenAI pages cover eligibility and common rejection reasons.
  2. 2

    Add the API keys to Cursor

    In Cursor settings, under Models, paste your Anthropic and OpenAI keys and enable them. Model calls now bill to those accounts and draw on the credits instead of your Cursor allowance.
  3. 3

    Put Cursor in Auto mode by default

    Auto picks a cheaper model when the task allows and only escalates to Opus-class models when it needs to. Override per task, not globally.
  4. 4

    Get the team onto GitHub for Startups

    Free GitHub Team seats, Actions minutes and Codespaces credits for eligible startups. If some engineers do not need Cursor daily, GitHub Copilot on that plan is a cheaper second seat.
  5. 5

    Right-size seats quarterly

    Cursor Teams analytics show per-user agent usage. Move heavy users to Premium seats and light users to Standard, and drop seats for anyone who has not opened the editor in a month.

Cursor vs Claude Code vs GitHub Copilot: the math

For a three-engineer startup doing agentic coding daily, here is roughly what a year costs on each route. Model usage assumes you are on startup credits where the tool allows it.

RouteYear-one cost (3 seats)What it coversWhere it breaks
Cursor Pro x3 $720 Full editor plus agent for three people Pro limits run out on heavy days; upgrade path is Pro+ or Ultra per seat.
Cursor Teams Standard x3 $1,440 Same as Pro plus SSO, admin, privacy mode Double the per-seat price for the admin layer. Same limit ceiling.
Cursor Ultra x3 About $7,200 Effectively unlimited agent usage Expensive if only one engineer is a heavy user. Mix tiers instead.
Claude Code on Max 5x x3 $3,600 Terminal agent with Claude usage included No editor. Pair with any IDE. See Claude Code pricing.
Cursor Pro x3 + startup API credits $720 + $0 model usage Editor plus Claude and GPT model calls billed to credits Credits expire. Some Cursor features still count against plan limits.
GitHub Copilot via GitHub for Startups $0 (eligible startups) Inline completions and chat inside VS Code or JetBrains Weaker agent mode than Cursor for large refactors.

The right stack for most funded startups: Cursor Pro or Teams Standard seats, Anthropic and OpenAI keys on startup credits, and GitHub for Startups for the repo, CI and a Copilot seat for anyone who does not need Cursor.

Questions founders ask

How much does Cursor cost per month?
Cursor Hobby is free. Cursor Pro is $20 a month. Pro+ and Ultra are higher-priced tiers with 3x and 20x Pro's agent limits respectively. Teams Standard is $40 per user per month, Teams Premium has 5x the limits, and Enterprise is custom pricing.
Is Cursor free?
There is a free Hobby plan with no credit card required. It includes the full editor and a limited number of agent requests, which is enough to evaluate Cursor but not enough to use it as your daily driver. Real daily use needs Pro at $20 a month.
What is the difference between Cursor Pro and Pro+?
Pro+ includes everything in Pro with three times the agent limits and higher Grok Bot usage. It exists for developers who hit Pro's usage ceiling most days. Ultra goes further with twenty times Pro's limits and priority access to new features.
Does Cursor have a student discount?
Not a standing one as of our last check. Cursor's student page points to promotions at campus and online events rather than a fixed offer. Cursor previously ran a free year of Pro for verified students, so check cursor.com/students for the current status.
Does Cursor have a startup program?
No. Cursor does not run a startup credit program. Startups reduce their bill by bringing their own Anthropic and OpenAI API keys, which can be funded by Claude for Startups and OpenAI for Startups credits, and by using GitHub for Startups for the rest of the engineering stack.
Can I use my own API key in Cursor?
Yes. Cursor lets you add Anthropic, OpenAI and other provider keys in settings. Model calls then bill to those accounts instead of counting against your Cursor plan limits. Some features that depend on Cursor's own model routing will still use your plan allowance.
Is Cursor Ultra worth $200?
Only if you run agents most of the day and hit Pro+ limits regularly. For terminal-driven work, Claude Code on Max 20x costs the same and includes Claude model usage. Many teams put one heavy user on Ultra and everyone else on Pro.

Related guides

Programs in this guide