Cursor pricing explained: plans, limits, and is it worth it? (2026)

Cursor pricing in 2026 runs on five main tiers. Hobby is free, Pro is $20 per month, Pro+ is $60 per month, Ultra is $200 per month, and Teams starts at $40 per user per month, with Enterprise priced on request. Annual billing takes roughly 20 percent off, so Pro drops to about $16 per month and Teams to about $32 per user. The short version of who each plan is for: Hobby suits people trying the editor, Pro fits most working developers, Pro+ and Ultra are for heavy agent users who burn through model usage, and Teams and Enterprise add billing, SSO, and admin controls for organizations.

I run engineering at ProductOS, and my team has paid for Cursor across several of these tiers. The tricky part is not the sticker price. It is understanding how the usage credits actually get spent, because that is where a $20 plan can quietly turn into a much bigger bill. Here is the full breakdown, verified against Cursor’s official pricing page.

Cursor pricing at a glance

How Cursor pricing works

The most important thing to understand about Cursor pricing is that it is not a flat all you can eat subscription. Since 2025, Cursor meters your agent and chat usage against a pool of included model usage that is denominated in dollars. Your monthly fee buys you a matching pool of compute. Pro gives you about $20 of included usage, Pro+ gives roughly three times that, and Ultra gives roughly twenty times that.

Every time you send an agent request or a chat message, you draw down that pool. How fast you draw it down depends entirely on which model you pick and how much context you feed it. A quick autocomplete is close to free. A large refactor across a big repository using a frontier model like Claude Opus or GPT-5 can consume a meaningful chunk of your monthly pool in a single afternoon. Two levers matter most here. Auto mode lets Cursor pick the cheapest capable model, which stretches your credits. Max mode expands the context window and bills at token rates, which is powerful but expensive.

One thing worth separating out. Tab completions, the inline autocomplete that Cursor is famous for, are handled generously across paid plans and are not the thing that blows up your bill. It is the agent and chat calls to premium models that eat the pool. If you want a broader picture of how these tools are structured, our overview of AI coding tools covers the landscape.

The plans in detail

Hobby (free)

The free tier gives you a limited number of agent requests and limited Tab completions, with no credit card required. It is genuinely useful for evaluating whether Cursor fits your workflow. It is not enough for daily professional work, and you will hit the ceiling quickly if you lean on the agent.

Pro ($20 per month)

Pro is the plan most individual developers should start on. You get roughly $20 of included model usage per month, extended agent limits, access to frontier models, and the full feature set including MCPs, skills, hooks, and cloud agents. Billed annually it works out to about $16 per month, or roughly $192 per year. For most people writing code a few hours a day with sensible model choices, the included pool is enough.

Pro+ ($60 per month)

Pro+ is Pro with about three times the included usage. If you are running the agent all day and repeatedly hitting your Pro pool before month end, Pro+ is cheaper than paying overage rates on top of Pro. This is the plan for people who have made agentic coding their default way of working.

Ultra ($200 per month)

Ultra gives roughly twenty times the included usage of Pro plus priority access to new features. This is aimed at power users who run large, long-context, agent-heavy workflows constantly. If you are a professional who would otherwise blow past Pro+ every month, the math can favor Ultra. For most people it is overkill.

Teams ($40 per user per month)

Teams bundles the individual feature set with the things organizations actually need: centralized billing, admin controls, usage analytics, team-wide privacy mode, and SAML or OIDC SSO. Billed annually it drops to about $32 per user. There is also a premium team tier with several times the standard included usage for heavier teams.

Enterprise (custom)

Enterprise adds pooled usage across your whole organization, SCIM seat management, repository and model access controls, audit logs, an AI code tracking API, and priority support. Pricing is negotiated. This is for larger companies that need governance and procurement to sign off.

The usage-limit gotchas

This is the section I wish someone had spelled out for me. Cursor’s move to usage-based pricing was not smooth. In June 2025 Cursor changed the Pro plan from a predictable model of around 500 fast requests per month to the dollar-denominated usage pool it uses today. Many users felt blindsided, reported running out of usage after only a handful of prompts on the newest Claude models, and saw effective costs jump sharply for agentic work. Cursor publicly apologized in July 2025, acknowledged that the rollout was handled poorly, and offered refunds for surprise charges. You can read the contemporaneous account on TechCrunch.

The practical gotchas that remain in 2026:

• Model choice is your biggest cost lever. A frontier model can cost many times what a lightweight model costs for the same task. Leave Auto mode on unless you have a reason not to.
• Max mode is a trap if you forget it is on. The expanded context window bills at token rates and can consume a large share of your pool in one refactor.
• Included usage is not unlimited requests. Once you exhaust the pool, on-demand usage continues and is billed after the fact. Set a spend limit so you are not surprised.
• Big repositories cost more. Referencing many files at once inflates context, and context is what you pay for.

None of this makes Cursor a bad deal. It just means the honest way to think about Cursor pricing is as a metered utility with a subscription baked in, not a fixed monthly cost.

Is Cursor worth it?

For a working developer, yes, Pro at $20 is easy to justify. The Tab completions alone save meaningful time, and the agent is genuinely capable for real tasks. The question is not whether Cursor is worth $20. It is whether your usage stays inside the pool or spills into overage, because that is what determines your real monthly cost.

My honest guidance: start on Pro, turn on Auto mode, and watch your usage for two or three weeks. If you consistently exhaust the pool before month end, do the math on Pro+ or Ultra rather than paying overage on Pro. If you barely touch it, the free tier or Pro is all you need. And if predictable billing matters more than raw model access, that is exactly where the alternatives get interesting. We go deeper on the workflow side in our guide to vibe coding.

Cursor pricing vs alternatives

Context helps. GitHub Copilot sits lower on price, with a free tier, Pro at $10 per month, Pro+ at $39, and Business at $19 per user, though it moved to its own AI credits model in 2026. Cursor tends to feel more capable for agentic, multi-file work, which is part of why it costs more. Tools like Claude Code and Windsurf compete on the same agentic ground with their own billing quirks.

If your goal is shipping a working application rather than editing code line by line, the comparison changes shape entirely. Our roundup of the best AI app builders covers tools that price by output rather than by metered compute, and our ProductOS vs Cursor comparison lays out where each approach wins. If you want to try building something end to end, you can start building on ProductOS.

Frequently asked questions