Cursor Pricing Explained (2026): Every Plan Compared

Cursor's pricing runs two ladders: an individual one — free Hobby, Pro at $20/month, then Pro+ at $60/month and Ultra at $200/month for heavier agent use — and a team one, starting at Teams from $40/user/month and topping out at custom-quoted Enterprise. That structure is straightforward on the page, but the real question most developers wrestle with is whether the jump from free to Pro is worth it, and whether a team plan earns its per-seat premium over everyone just running Pro individually. Pricing pages rarely tell the full story. The interaction between model selection, agent usage allowances, and cloud compute makes Cursor's billing more layered than a flat monthly fee suggests.

Plans at a Glance

India-based developers also have access to a regional Start plan (₹649/month, tax-inclusive) with a narrower scope than Pro — Cursor Models pool access only, without the Other Models pool, Bugbot, Auto, or Automations (per Cursor's Models & Pricing docs, checked August 2026). It's not a global entry-level tier, so it's left out of the table above.

What Is Cursor and Why Does Pricing Matter?

Cursor is a VS Code fork rebuilt around AI-native coding. It ships with deep integration of large language models for code completion, multi-file editing, chat, and agentic workflows that can plan and execute multi-step tasks across a codebase. If you've been following the AI code editor space, you'll know it sits alongside GitHub Copilot and Windsurf (formerly Codeium) as one of the three tools most developers are evaluating right now.

Pricing matters here more than it does for a typical SaaS subscription because your effective monthly cost isn't just the plan fee. It's the plan fee plus the inference cost of whichever models you choose to run, multiplied by how heavily you use agent mode. Two developers on the same Pro plan can end up with meaningfully different experiences depending on whether they're running lightweight completions or spinning up Cloud Agents on demanding tasks. The headline number on the pricing page is a starting point, not a ceiling.

What Do You Actually Get on Each Plan?

Hobby (Free)

The Hobby tier is a trial lane. You get access to Cursor's editor, basic AI completions, and a taste of the agent capabilities. The model selection is restricted, and usage caps are tight enough that anyone coding for more than a few hours a day will hit them. It's designed to let you feel the difference between Cursor and a vanilla VS Code setup, not to be a permanent home.

That said, for weekend projects or developers who just want to see what the fuss is about, it does the job. No credit card required. No time limit. Just a clear ceiling on what you can do.

Pro ($20/month)

This is where Cursor becomes a genuine daily driver. Per Cursor's pricing page (checked August 2026), the Pro plan is listed at $20/month and unlocks the full model roster, a substantially larger agent usage allowance, and access to Cloud Agents. The agent usage allowance scales with your plan but is also shaped by the inference cost of whichever model you select for a given task; running a frontier-class model eats through your allowance faster than a smaller, cheaper one.

Pro carries the same Privacy Mode toggle as the free Hobby tier — you can control whether your code is retained for training or stored beyond the immediate session, per Cursor's security documentation. For freelancers working with client codebases, that toggle matters regardless of which individual plan you're on.

Pro+ and Ultra (higher individual limits)

Per Cursor's Models & Pricing docs (checked August 2026), Pro+ is $60/month and includes $70/month of third-party model usage — more usage than the fee itself, alongside 3x Pro's agent allowance. Ultra is $200/month and includes $400/month of model usage, alongside 20x Pro's allowance. Both prices are published outright, not custom or usage-billed after the fact. They're aimed at developers who lean hard on agent mode and keep bumping into Pro's allowance, not at teams.

Teams (from $40/user/month) and Enterprise

Per Cursor's pricing page (checked August 2026), Teams starts at $40/user/month across two tiers — Standard and Premium, the latter offering 5x Standard's agent limits. The feature delta over the individual plans is organisational control: centralised billing, admin dashboards, the ability to enforce privacy mode across every seat, and SAML/OIDC SSO. Enterprise sits above Teams on a custom quote, adding pooled usage, SCIM seat management, audit logs, and invoice billing.

If you're a solo developer or a two-person startup, Teams is overkill — Pro or Pro+ covers you individually. But for a team of six or more where someone needs to manage seats, enforce data handling policies, or integrate with SSO, the admin layer justifies the premium. The per-seat cost is steep at scale. A 20-person team on Teams Standard is looking at $800/month (around £640) before anyone considers model inference overages. That's the kind of line item that needs sign-off, not a quick expense claim.

How Does Agent Usage Affect Your Bill?

This is the part most pricing breakdowns skip, and it's arguably the most important.

Cursor's agent mode lets the AI plan and execute multi-step coding tasks: reading files, writing code, running terminal commands, iterating on errors. It's the feature that separates Cursor from a simple autocomplete overlay. But agent usage isn't unlimited on any plan. Each plan comes with an allowance, and the rate at which you consume that allowance depends on two things: how many agent sessions you run, and which underlying model powers them.

Choosing a top-tier frontier model for an agent task will burn through your allowance considerably faster than choosing a smaller, cheaper model. Cursor's UI typically shows you which model you're using and gives some indication of the cost implication, but it's not always obvious until you've been through a billing cycle or two. Developers who lean heavily on agentic workflows, the kind where you describe a feature and let the agent scaffold it across multiple files, should expect to monitor their usage dashboard regularly.

The advertised plan price is a floor, not a fixed cost. Agent-heavy workflows on frontier models can push effective monthly spend well above the base subscription, especially for Pro users near their allowance limits. Cursor's own pricing page notes that usage depends on plan allowance and model inference cost.

Cloud Agents: a new variable

Cursor Cloud Agents run in remote, internet-enabled virtual machines. They can clone your repository, execute commands, install dependencies, and push changes, all without touching your local machine. It's powerful. It's also a surface area expansion worth understanding.

Per Cursor's security documentation, Cloud Agents operate in sandboxed VMs with repository access scoped to the task. But "internet-enabled" and "auto-run commands" in the same sentence should make any security-conscious team pause. Questions worth asking before enabling Cloud Agents for your team: What repository data is retained after a session ends? How are secrets and environment variables handled? What's the egress policy? And what mitigations exist against prompt-injection attacks that could cause an agent to execute unintended commands?

These aren't hypothetical concerns. They're the kind of thing a responsible engineering lead should review against their own threat model before rolling Cloud Agents out across a team. Cursor's docs address some of these; others require a conversation with their sales or security team, especially if you're subject to UK GDPR or handling regulated data.

Privacy, Security, and What Gets Stored

Cursor's security page documents several things that matter for privacy-conscious developers. The editor performs codebase indexing to power its context-aware suggestions, which means embeddings or metadata derived from your code are stored (at least transiently) on Cursor's infrastructure. Subprocessors handle parts of this pipeline.

Privacy Mode is available on the free Hobby tier and Pro alike — any individual can enable it in settings. On Teams and Enterprise, admins can additionally enable or enforce it organisation-wide, which is the main reason compliance-heavy teams gravitate to the higher tiers. But "available to toggle" and "enforced by default" are doing different work. A solo Pro user who forgets to enable privacy mode is in a different position from a Teams org where the admin has locked it on for everyone.

For developers working under client NDAs, in regulated industries, or simply building proprietary products they'd rather not share with a third party's model training pipeline, it's worth reading the current data retention and subprocessor disclosures carefully. These policies evolve, and what was true six months ago may not be true today. Check the docs, not a cached blog post.

Which Plan Is Right for a Solo Developer vs. a Team?

Solo developers and freelancers

Pro is the obvious pick. The $20/month fee is comparable to what you'd pay for GitHub Copilot's individual plan, and Cursor's agent capabilities go meaningfully further than inline autocomplete. If you're evaluating Cursor against Copilot, our Cursor vs. GitHub Copilot comparison breaks down the workflow differences in detail.

Freelancers should pay particular attention to the privacy toggle. Client work deserves a clear data handling posture, and Pro gives you the tools to set one, as long as you actually use them.

Small teams (2–10 people)

This is the awkward middle ground. A team of four on Teams Standard is $160/month. That's not trivial, but it's also not enterprise money. The question is whether centralised admin and enforced privacy mode are must-haves or nice-to-haves. If everyone on the team is technical enough to enable their own privacy settings and you trust them to do it, individual Pro or Pro+ seats might be sufficient. If you need an audit trail or SSO, you need Teams — or Enterprise if you also need pooled usage and SCIM.

Larger teams and enterprises

Teams or Enterprise is the only realistic option, and even then, teams dealing with sensitive codebases should conduct a thorough review of Cursor's subprocessor list, data residency, and Cloud Agent architecture before committing. Enterprise earns its extra step once you need pooled usage, SCIM provisioning, or audit logs beyond what Teams' per-seat billing comfortably covers. The best Cursor alternatives guide covers options for teams that need tighter self-hosted or air-gapped setups.

Final Verdict

Cursor Pro at $20/month is the right plan for most developers. It unlocks the full AI model roster, gives you enough agent usage for serious daily work, and includes the same privacy controls as the free tier. The free Hobby tier is a test drive, nothing more; Pro+ ($60/month) and Ultra ($200/month) are there if you personally need more agent headroom than Pro's allowance covers. Teams (from $40/user/month) justifies its price tag only when you need organisational controls, enforced compliance settings, or are scaling past a handful of seats — with Enterprise above it for pooled usage and audit requirements.

The biggest gotcha across all plans is that agent usage and model selection create variable costs on top of the base subscription. Developers who rely heavily on agentic workflows with frontier models should monitor their usage closely, especially in the first few billing cycles.

Best for: Solo developers and small teams who want AI-assisted coding that goes well beyond autocomplete, with enough flexibility to choose models and control data handling.

Avoid if: Your security requirements demand fully self-hosted inference, air-gapped environments, or you're uncomfortable with codebase metadata being processed on third-party infrastructure, even with privacy mode enabled.

Frequently Asked Questions

Is Cursor free to use?

Yes, Cursor offers a free Hobby plan with limited model access and capped agent usage. It's suitable for evaluation but not for daily professional work.

How much does Cursor Pro cost per month?

Cursor Pro is listed at $20/month (approximately £16) as of August 2026. Actual costs may be higher if you consume agent usage beyond your plan's included allowance.

What's the difference between Cursor's individual and team plans?

Cursor's Pro+ ($60/month) and Ultra ($200/month) tiers give individuals higher agent-usage allowances — 3x and 20x Pro's limits respectively, each with a published price, not a custom quote. Teams (from $40/user/month) adds centralised billing, admin dashboards, and organisation-wide privacy enforcement; Enterprise adds pooled usage, SCIM, and audit logs on top of that.

Does Cursor store my code?

Cursor performs codebase indexing and stores embeddings or metadata to power context-aware suggestions, per its security documentation. Privacy Mode is available on the free Hobby tier and Pro for individuals to enable themselves, and can additionally be enabled or enforced organisation-wide by Teams and Enterprise admins. Review Cursor's current security page and subprocessor list for the latest details.

Are Cursor Cloud Agents safe to use with private repositories?

Cloud Agents run in remote internet-enabled VMs with repository access and auto-execute commands. Before using them with private or sensitive code, review Cursor's documentation on session retention, secrets handling, egress policies, and prompt-injection mitigations. Teams subject to UK GDPR or similar regulations should assess this against their own threat model.

How does Cursor compare to GitHub Copilot on pricing?

Both offer individual plans at a similar price point. Cursor's agent mode and multi-file editing capabilities go further than Copilot's inline suggestions, but the right choice depends on your workflow. Our Cursor vs. GitHub Copilot breakdown covers the differences beyond pricing.