Labs · Review2026 Edition

Cursor Review 2026

Cursor is still the most complete AI code editor we've reviewed: a VS Code-based editor, an agent that edits across many files at once, cloud agents that keep working while you're away, and Bugbot reviewing your pull requests. But the deal changed a lot in 2026. SpaceX now owns Cursor, and new customers buy it through the X Subscription, on a new ladder: Free at $0, Plus at $8, Premium at $30, Super at $100 and Ultra at $200 a month.

Here's the catch: Claude, GPT and Gemini are no longer included. You pay for them on top, at API rates, from the very first request. In this Cursor review, we break down what each plan really gets you, whether to keep your legacy $20 Pro, the pain points developers keep raising, practical tips to keep the bill in check, and the alternatives worth a look. Our score: 4.0/5.

Verdict · 5 criteria scored

Our Cursor review in a nutshell

Romain Cochard
Reviewed by
Romain Cochard
CEO

Cursor still earns its reputation. No other editor weaves AI into everyday coding this well. You keep VS Code, your extensions and your keybindings. The agent reads your codebase, proposes changes across files and waits for your approval. Cloud agents and Bugbot carry the work all the way to the pull request.

What slipped is the value. For a new subscriber, the first serious plan jumps from $20 to $30 a month, and Claude or GPT now cost extra. The included allowance resets weekly, it's shared with the Grok apps, and its size is only published as multiples ("4x Plus"). Hence our 4.0/5: a great tool, with a pricing model you need to read closely before you commit.

Free trial

The numbers speak. Want to try Cursor?

Try Cursor for free→
What you actually pay

Cursor pricing under the X Subscription

Since 2026, every new subscriber goes through the X Subscription, which bundles Cursor, Grok and X Premium. Each plan comes with a weekly usage allowance on Cursor's own models (Grok and Composer). Third-party models like Claude, GPT and Gemini are never included: from Premium up, you buy them separately at API rates from a prepaid balance. The legacy plans (Hobby, Pro, Pro+, Ultra, Teams) still work, but only for people who already had them.

Monthly prices in USD, excluding sales tax and VAT

Free

To try the editor

$0/mo

Limited agent and chat

  • Limited Agent and Chat
  • The full editor, VS Code extensions included
  • Privacy Mode available
  • No third-party models

Plus

For coding with in-house models

$8/mo

$80 per year

  • Grok 4.7, 4.6, 4.5 and Composer 2.5
  • Fixed settings (Grok 4.7 at 256k context)
  • No Auto, cloud agents or Bugbot
  • No third-party models or your own API keys
Best starting point

Premium

Cursor's pick for new Agent users

$30/mo

$300 per year

  • 4x the usage of Plus
  • Cloud agents, automations and Bugbot
  • Third-party models at API rates, on top
  • Origin code hosting and Grok Bot

Super

For daily heavy use

$100/mo

$1,000 per year

  • 15x the usage of Plus
  • Everything in Premium
  • Priority access at peak hours
  • Third-party models still extra

Ultra

For power users

$200/mo

$2,000 per year

  • 50x the usage of Plus
  • Everything in Super
  • Early access to new features
  • Third-party models still extra

Teams

For engineering teams

From $30/seat/mo

Seats at $30, $100 or $200

  • SAML or OIDC SSO
  • Centralized billing
  • Privacy Mode enforced team-wide
  • Extra $0.25 per million tokens on third-party models

Enterprise is custom-priced: pooled usage, SCIM, audit logs, priority support and invoice billing. Regional data residency adds 10% to the price of eligible models. To avoid paying for the next period, cancel at least 24 hours before renewal.

Community · developer reviews

What developers really say about Cursor

The Hack'celeration verdict

We compared Cursor on five criteria.

Ease of use, real cost, features, support and integrations, against the other AI code editors on the market.

Criterion 01 · Ease of use

How easy is Cursor to pick up?

4.5/5

Very easy, especially if you come from VS Code. Cursor is a fork of it: on first launch it imports your extensions, theme and keybindings. You don't relearn your tool, you add an AI layer on top. That's its biggest edge over tools that force you into a new environment.

Day to day, three moves cover most of it. Tab accepts the grayed-out suggestion. Cmd+I (Ctrl+I on Windows and Linux) opens the agent: describe what you want in plain English, and it reads the relevant files, proposes edits and shows you every diff before applying it. Chat answers questions about your code without touching anything.

Two things made onboarding rougher in 2026. First, Cmd+K, the quick inline edit and terminal command generator, was retired and folded into the agent. Plenty of regulars find the agent slower for small tweaks. Second, picking a plan and tracking the weekly allowance takes real reading time. The editor stayed simple. The pricing didn't.

Why 4.5? A developer who knows VS Code is productive within an hour, and a beginner quickly gets the "ask, review, accept" loop. We knock off half a point for losing Cmd+K and for usage settings that aren't obvious.

Verdict
The easiest AI editor for anyone coming from VS Code, as long as you read the pricing before you pay.
Criterion 02 · Value for money

What does Cursor really cost in 2026?

3.0/5

It depends on two things: which model you code with, and when you subscribed. Sign up today and you go through the X Subscription. Plus at $8 a month gets you the in-house models (Grok 4.7, 4.6, 4.5 and Composer 2.5), but no Auto, no cloud agents, no Bugbot and no Claude. Premium at $30 is the first complete plan, with 4x the usage of Plus. Super at $100 gives 15x, Ultra at $200 gives 50x.

The big shift is third-party models. The legacy $20 Pro included an allowance for Claude, GPT and Gemini. The X plans drop it: you pay for every request at API rates from a prepaid balance, even on Ultra. For reference, Claude Sonnet 5.5 runs $2 per million input tokens and $10 per million output tokens; Claude Opus 5.5 runs $4 and $20. A big codebase burns through millions of tokens quickly. By contrast, in-house models beyond your allowance cost $2 and $6 for Grok 4.7, and $0.50 and $2.50 for Composer 2.5.

The allowance resets weekly, and it's shared across the editor agent, the CLI, cloud agents, Grok Bot and the Grok apps. A Grok chat that has nothing to do with code eats into the same pool. Its size isn't published in dollars or tokens, only as multiples of Plus. Cursor says every paid X plan includes at least twice the coding usage on its own models compared to the legacy plan it replaces. That holds for in-house models, but the third-party allowance is simply gone.

Why 3.0? Plus at $8 is one of the cheapest ways to code with an agent. But the moment you want Claude or GPT, Cursor costs more than GitHub Copilot Pro at $10 (with $15 of AI credits included) or Claude Code, which comes with a $20 Claude Pro plan. And a bill you can't forecast is a real problem for freelancers.

Verdict
A bargain on in-house models, a bill that climbs fast if you code with Claude or GPT.
Criterion 03 · Features and depth

What can Cursor do that others can't?

4.5/5

The agent is the heart of the product. Hand it a task ("add a date filter to this list and update the specs") and it reads the files involved, writes the code, runs terminal commands and fixes its own mistakes. You approve every change. For longer jobs, cloud agents do the same work on a remote machine: close your laptop, come back to a branch ready for review. Since 2026 you can even monitor and steer your agents from the iOS app.

Bugbot reviews your pull requests on GitHub or GitLab. It flags bugs, security issues and quality problems, and every comment comes with a "Fix in Cursor" link. It runs on each PR update, or on demand. Automations, the Agent SDK and the Origin app marketplace round things out for teams that want to scale this up.

Rules are the most underrated part. You write your conventions in .mdc files under .cursor/rules, version-controlled with your code, or in a plain AGENTS.md at the root. The agent reads them on every task: naming, libraries to avoid, how specs are structured. Teams also get shared rules for every member.

On models, Cursor offers the widest lineup we've seen: Claude Sonnet 5.5 and Opus 5.5, GPT-5.6, Gemini 3.1 Pro, Grok 4.7 and Composer 2.5, plus a router that picks for you. Why not a 5? The best of those models now cost extra, and the agent can still be confidently wrong. Reviewing its output is still your job.

Verdict
The most complete AI editor, from the first line of code to a reviewed pull request.
Free trial

Sold on the details? Start a Cursor trial.

Try Cursor for free→
Criterion 04 · Customer support

Can you count on Cursor's support?

3.0/5

For how-to questions, yes. The official forum is busy and Cursor staff often answer directly, including on touchy topics like the move to the X Subscription. The docs cover plans, rules, Bugbot and privacy well. That's where you'll get answers quickest.

For account or billing issues, it's a different story. Public customer reviews complain about email support that answers with an AI first, refunds that are hard to get and a few account suspensions with little explanation. One legacy Pro subscriber also says they saw the jump from $20 to $30 without a warning email. And a bug hit some legacy plans: the "other models" pool showed 100% used with no usage once the in-house pool ran out.

History matters too. In 2025, Cursor had already overhauled its billing (from requests to a credit pool), then publicly apologized and refunded unexpected charges. The 2026 change is better documented, but switching is still permanent.

Why 3.0? A responsive forum and clear docs, but human support is too hard to reach for a tool that can cost $200 a month. Priority support is reserved for Enterprise.

Verdict
Good for how-to questions, weak as soon as money is involved.
Criterion 05 · Integrations

Does Cursor plug into your stack?

4.5/5

Yes, into almost everything a dev team uses. Code hosts: GitHub, GitLab, Bitbucket and Azure DevOps, for Bugbot and cloud agents. Team chat: Slack and Microsoft Teams, where you can hand a task to an agent from a message. Issue tracking: Jira and Linear, to go from a ticket to a branch. Notion covers the docs side.

Cursor also speaks MCP (Model Context Protocol), the standard for connecting the agent to a database, an internal tool or a third-party service. Add an MCP server and the agent can read your schema or open a ticket without leaving the editor. If you'd rather stay in your current environment, there's a JetBrains integration and Xcode support, plus a CLI to run the agent in a terminal or a script.

Security is a strong point for companies. Cursor holds AIUC-1, ISO/IEC 27001:2022 and ISO/IEC 42001:2023 certifications, plus a SOC 2 Type II report. Privacy Mode, available even on Free, means your code is never used to train models; it's on by default for Enterprise, and admins can enforce it for a whole team. Cursor runs no infrastructure in China and commissions a third-party penetration test at least once a year. A DPA is available, but there's no announced EU hosting: check that with your legal team if your data requires it.

Why 4.5? The catalog covers the essentials and MCP opens up the rest. Half a point off for the lack of EU hosting and for regional data residency costing 10% more.

Verdict
Connected to the whole dev workflow, with solid security credentials but no EU hosting.
Subscribed before the X Subscription?

Keep your legacy plan or switch

If you were paying for Cursor before the X Subscription, your plan still works. Switching to the new lineup is optional, but permanent: there's no way back. Cursor has also said legacy plans will be migrated at some point, with advance notice. Here's how to decide, based on what you have today.

01

You're on the legacy $20 Pro

Most common case

Its closest match is Premium at $30. You'd get at least twice the in-house usage, but you'd lose the Claude, GPT and Gemini allowance your Pro includes. If you mostly code with Claude, stay put as long as the plan exists. If you mostly use Composer or Auto, switching can make sense.

Worth knowing: annual Pro works out to $16 a month, roughly half of Premium billed monthly.

Closest match
Premium, $30/mo
You lose
Included third-party models
Our take
Stay if you code with Claude
02

You're on the legacy $60 Pro+

Its match is Super at $100. That's a steep jump ($40 more a month), and once again the third-party allowance goes away. Before switching, check in the editor settings how much of your usage goes through Claude or GPT. If it's a big share, your real extra cost will be well above $40.

Closest match
Super, $100/mo
Increase
$40/mo
Our take
Stay, unless you're 100% in-house
03

You're on the legacy $200 Ultra

The price doesn't change: Ultra stays at $200. You gain in-house usage and early access, but you lose the included third-party models. The math depends entirely on how much Claude and GPT you use. If you've already burned through your quota in a few days, the new weekly cap may block you more often.

Closest match
Ultra, $200/mo
You gain
In-house usage, early access
You lose
Included third-party models
04

Your team is on legacy Teams

Teams Standard at $40 per user maps to a $30 Premium seat, and Teams Premium at $120 maps to a $100 Super seat. On paper, the team pays less per seat. But on Teams, every third-party model request adds $0.25 per million tokens on top of API rates. Run the numbers on last month's real usage before you decide.

Standard $40
Premium seat at $30
Premium $120
Super seat at $100
On top
$0.25 per million third-party tokens
What the forums know

Eight tips to spend less and ship more with Cursor

These settings come from the official docs and from developers who use Cursor every day. They target the three most common bad surprises: a weekly allowance gone in two days, a third-party model bill that balloons, and an agent that ignores your instructions.

01

Set a spend cap before you turn on Auto Top-up

Biggest win

In billing settings, set your monthly additional usage limit first, and only then enable Auto Top-up. The other way round, one long agent session on Opus can refill your balance several times before you notice. Cursor says requests are never downgraded in quality or speed. That's exactly why they cost what they cost.

Where
Billing settings
Order
Cap, then top-up
02

Plan with a big model, build with Composer

A lot of developers use a frontier model to write the plan for a task, then Composer 2.5 to carry it out. Beyond the allowance, Composer costs $0.50 per million input tokens and $2.50 per million output tokens, versus $2 and $6 for Grok 4.7 and $4 and $20 for Claude Opus 5.5. On a long task, the gap adds up to real dollars.

Composer 2.5
$0.50 / $2.50 per million
Grok 4.7
$2 / $6 per million
03

Don't chat with Grok on your coding account

The weekly allowance is shared by the agent, the CLI, cloud agents, Grok Bot and the Grok apps. A single Grok conversation with zero code in it can eat several percent of your week. If you use Grok for other things, keep a separate account.

Allowance
Weekly and shared
Track it in
Editor settings
04

Keep your rules short and split up

When the agent's context overflows, long rule files are the first thing to get dropped. Several small .mdc files in .cursor/rules, each scoped to a file type, beat one giant catch-all file. For a small project, a plain AGENTS.md at the root is enough.

Folder
.cursor/rules
Simpler option
AGENTS.md
05

Exclude noise with .cursorignore

A .cursorignore file at the root, using .gitignore syntax, stops Cursor from indexing node_modules, dist or generated files. The agent then searches your real code, and you stop sending the model files that inflate token usage.

File
.cursorignore
Syntax
Same as .gitignore
06

Trigger Bugbot on demand

Bugbot runs on every PR update, but you can also kick it off whenever you like by commenting bugbot run on the pull request, or with cursor review from your terminal. Handy for one last pass on a branch right before merging.

Command
bugbot run
Plan
Premium and up
07

Replace Cmd+K with a tightly scoped agent

Cmd+K is gone as of 2026. To get quick edits back, select the lines first, then open the agent with Cmd+I, and add a line to your AGENTS.md like "for a change under ten lines, don't touch any other file". The agent is still slower, but it stays in its lane.

Shortcut
Cmd+I or Ctrl+I
Trick
Select before you ask
08

Go annual only after a full month

Annual billing saves about two months ($300 instead of $360 for Premium). But you can only judge a weekly allowance by real usage. Pay monthly for one month, note which day of the week you hit the cap, then choose between Premium, Super or annual.

Premium annual
$300/yr
Premium monthly
$360 over a year
Cursor not the right fit?

Five Cursor alternatives worth a look

If the X Subscription or third-party model pricing puts you off, these five tools do better on one specific point: price, free model choice or open source. Each card also spells out its limit. The ranking is editorial: a sponsored placement is always labeled on its card, and no affiliate link ever changes a score. For the full ranking, see our pick of the best AI coding tools.

01

GitHub Copilot: the cheapest with credits included

Cheapest

Copilot is GitHub's coding assistant, and most teams already know it. Its Pro plan costs $10 a month and includes $15 of AI credits you can spend across several models. If you code with Claude or GPT, it's often the lowest bill on the market.

Its limit: the agent is less capable than Cursor's on large edits across a whole codebase, and credits go fast on the top models.

Price
Free $0, Pro $10, Pro+ $39, Max $100
Included
$15 of AI credits on Pro
Limit
Less capable agent
02

Claude Code: if Claude is your model

For Claude fans

Claude Code is Anthropic's coding agent, in your terminal or your editor. It comes with Claude Pro at $20 a month ($17 billed annually), whereas Cursor would charge you for Claude on top. If Claude is your go-to model, it's the most logical alternative. Read our Claude Code review.

Its limit: a single model provider, and a rawer experience than a full editor if you like watching the code appear.

Price
In Claude Pro at $20, Max from $100
Best for
Claude fans
Limit
One provider only
03

Windsurf: the closest thing to Cursor

Windsurf, now under the Devin brand at Cognition, follows the same idea: a code editor with a built-in agent. Its Pro plan is still $20 a month and there's a free tier. If you want the old Cursor Pro experience without the X Subscription, start here. Read our Windsurf review.

Its limit: it has changed owners and branding, so keep an eye on the roadmap.

Price
Free $0, Pro $20, Max $200
Teams
$80/mo + $40 per seat
Limit
New owner
04

Zed: bring your own keys

Zed is an independent code editor that accepts your own API keys or external agents. Personal is free, Pro is $10 a month with $5 of tokens included, Business is $30 per seat. You choose the model and pay the provider directly.

Its limit: it isn't VS Code, so your extensions don't come along, and the ecosystem is smaller.

Price
Personal $0, Pro $10, Business $30
Models
Your API keys
Limit
No VS Code extensions
05

Cline: open source, no subscription

Open source

Cline is a free, open-source VS Code extension. You plug in your own API keys (Anthropic, OpenAI or Google, for example) and pay only for what you use, with no middleman markup. If you want full control over costs, it's the most transparent option.

Its limit: no cloud agents or built-in PR review, and an API bill that can climb with no ceiling unless you set one.

Price
Free extension
You pay for
Model calls
Limit
No cloud agents

Do you make a tool in this category? Email us: hello@hackceleration.com

Who is it for?

Is Cursor right for you?

Three profiles, three answers. Then six checks before you subscribe.

Yes, if you code daily and are fine with in-house models

A developer or small team living in the editor all day, who wants an agent that can change a whole codebase and can work with Composer and Grok. Premium at $30 covers most needs, Super at $100 the heaviest days. If you want help rolling it out, see our Cursor agency.

Be careful if Claude or GPT is your go-to model

Every third-party request is billed at API rates on top of your plan. If you still have the legacy Pro, keep it. If you're starting fresh, compare with Claude Code or Copilot first, since their credits are included.

No, if your budget is fixed and tight

A freelancer who needs to forecast every dollar will struggle with an unstated weekly allowance and pay-as-you-go third-party models. Plus at $8 is still an option if you stick to in-house models. To learn how to keep usage in check, there's also our Cursor training.

  • Have you checked which model you code with most?
  • Have you set an additional usage cap before enabling top-ups?
  • On a legacy plan, have you worked out what you lose by switching?
  • Is Privacy Mode on, or enforced by your admin?
  • Has your legal team signed off on the lack of EU hosting?
  • Have you added a .cursorignore and short rules to your project?
FAQ · 8 questions

Frequently asked questions

  • How much does Cursor cost now?

    Between $0 and $200 a month for individuals. Free is $0, Plus is $8, Premium $30, Super $100 and Ultra $200, before tax. Teams start at $30 per seat, and Enterprise is custom-priced. Keep in mind none of these prices include Claude, GPT or Gemini, which you buy on top at API rates from Premium up. Your real budget is the plan plus your third-party model usage.

  • Is Cursor still worth it after the X Subscription?

    Yes, if you code every day and can work mostly with Composer and Grok: the editor and agent are still the best around, and you get at least twice the in-house usage of the old plans. Less so if you live in Claude or GPT, because every request is now billed on top. In that case, Copilot Pro at $10 or Claude Code in Claude Pro at $20 will usually cost you less.

  • Should I switch from legacy Pro to Premium?

    Only if you mostly use in-house models. Switching is permanent, and Premium costs $30 instead of $20 while dropping the third-party allowance your Pro includes. If you code with Claude, stay on legacy Pro as long as it's available. Cursor has said legacy plans will be migrated eventually, with advance notice, so you don't need to rush.

  • Are Claude and GPT included in Cursor plans?

    Not in the new plans. Third-party models are billed at API rates from the first request, from a prepaid balance, and only from Premium up. Claude Sonnet 5.5, for example, costs $2 per million input tokens and $10 per million output tokens. Plus at $8 gives you no third-party models at all, and no bring-your-own-key option either.

  • Cursor vs GitHub Copilot: which is cheaper for daily use?

    Copilot, in most cases. Copilot Pro is $10 a month with $15 of AI credits included, and it doesn't require the X Subscription. Cursor Premium is $30 with no Claude or GPT included. Cursor wins on agent depth, cloud agents and Bugbot. If you mainly want strong completions and occasional agent work with Claude, Copilot will cost you noticeably less.

  • Cursor vs Claude Code?

    It comes down to your favorite model. If you code with Claude, Claude Code is included in Claude Pro at $20, while Cursor bills Claude on top. If you want a full editor, multiple model providers and Bugbot on your pull requests, Cursor is richer. Nothing stops you from using both: Claude Code in the terminal, Cursor for review.

  • Does Cursor train on my code?

    Not if you turn on Privacy Mode. It's available on every plan, Free included, and guarantees that neither Cursor nor the model providers train on your code. It's on by default for Enterprise, and team admins can enforce it for everyone. Check it's enabled in settings on day one.

  • What happens when I hit the weekly limit?

    You have three options. Wait for the weekly reset. Switch to Composer 2.5, the cheapest model beyond the allowance at $0.50 per million input tokens. Or move up a plan, from Premium to Super. Before anything else, make sure Grok Bot or the Grok apps aren't drawing from the same pool: it's shared with your coding usage.

Hack'celeration Lab

Get our next reviews in your inbox

One email when a new tool review or comparison goes live: real prices, hidden limits and our verdict.

No spam. Unsubscribe anytime.