Cursor Review (2026): Still the Best AI Code Editor?
Cursor review: Agent, Tab, Cloud Agents and Bugbot, the new X Subscription prices from $8 to $200, which plan to pick and when Cursor is worth paying for.
PromptWises is reader-supported. Some links may earn us a commission at no extra cost to you. This never affects our ratings or recommendations.
Cursor is still worth it if you want an AI agent to do real, multi-file work inside your editor: it combines Agent, Tab autocomplete, Cloud Agents and Bugbot code review in one product, and nothing else bundles all four as well. Since 12 October 2026, new customers buy it as part of The X Subscription. Premium at $30 a month (at the time of writing) is the plan most developers need; Plus at $8 is too limited for daily coding. If you mainly want Claude or GPT models, budget for API costs on top.
A note on method: we have not completed hands-on testing for this review. It is based on Cursor’s pricing page, documentation, help centre and changelog, xAI’s X Subscription docs and OpenAI’s statements, first checked on 17 October 2026 and re-checked on 26 October 2026 (prices unchanged). The rating of 4.0 is provisional until we test it ourselves. If you are new to AI coding tools, start with our getting-started guide to AI coding assistants.
What Cursor is (and who owns it now)
Cursor is a desktop code editor built on VS Code, made by Anysphere. Your VS Code extensions, themes and keybindings carry over, and the AI features are built into the editor rather than bolted on as a plugin. The main pieces are:
- Agent: a side-panel agent that reads your codebase, edits files, runs terminal commands and searches the web, with Agent, Plan and Ask modes.
- Tab: autocomplete that suggests multi-line edits and predicts your next change, including in other files.
- Cloud Agents: agents that run in isolated cloud machines, work in parallel and open pull requests. You can start them from the editor, the web, the iOS app, Slack, GitHub comments or Linear.
- Bugbot: an automated reviewer for pull requests.
- Cursor CLI, an SDK, and a JetBrains route via the Agent Client Protocol (it needs a paid plan and JetBrains’ AI Assistant plugin).
Ownership changed this year. SpaceX completed its acquisition of Cursor on 14 August 2026 (Cursor’s announcement), and on 12 October Cursor’s own plans were replaced for new customers by The X Subscription, one plan covering Grok, Grok Bot, Cursor and X Premium. The editor still ships quickly: on 16 October the changelog added shareable canvases built by cloud agents and an Environments API in beta.
Cursor pricing in October 2026
Prices below are from Cursor’s models and pricing docs and xAI’s X Subscription pricing, at the time of writing (USD, excluding tax).
| Plan | Monthly | Annual | Included usage | Third-party models | Cloud Agents and Bugbot |
|---|---|---|---|---|---|
| Free | $0 | $0 | Limited | No | No |
| Plus | $8 | $80 | Base | No | No |
| Premium | $30 | $300 | 4x Plus | API list price, prepaid credit | Yes |
| Super | $100 | $1,000 | 15x Plus | API list price, prepaid credit | Yes, plus priority access at peak times |
| Ultra | $200 | $2,000 | 50x Plus | API list price, prepaid credit | Yes, plus earliest access to new features |
| Teams | $30 / $100 / $200 per seat | - | Matches seat tier | API list price plus Cursor Token Rate | Yes |
| Enterprise | Custom | - | Pooled | API list price plus Cursor Token Rate | Yes |
Three things matter more than the headline prices.
The included pool is weekly and covers only Cursor’s own models. It pays for Grok 4.7, 4.6 and 4.5 and Composer 2.5, resets every week and does not roll over. It is shared across everything on the subscription: Agent in the editor, the CLI, Cloud Agents, Grok Bot and the Grok apps. Heavy Grok chat use therefore comes out of your coding allowance.
Third-party models cost extra from the first request. Claude, GPT and Gemini are not in the pool on any plan. On Premium and above they draw on prepaid credit at API list price. Plus and Free cannot use them at all. A developer who runs most agent work on Claude Sonnet 5.5 pays $30 a month plus Sonnet’s usage at $2 / $10 per million input / output tokens. One who sticks to Composer 2.5 and Grok stays inside the pool until it runs out, then pays from credit.
Included usage is not published in absolute terms. Cursor gives only multiples (“4x Plus”), so you cannot tell in advance how much agent work a plan covers. xAI claims each plan includes “at least twice as much coding usage with our models as the comparable Cursor plan”; that is a vendor claim we cannot check.
Teams pay a Cursor Token Rate of $0.25 per million tokens on third-party requests, including requests made with their own API keys. Individual plans are exempt. More background is in our news report on the pricing change.
Should existing Cursor Pro users switch?
Probably not yet. Switching is optional, and legacy plans keep working until you change or cancel them (Cursor help). If you do switch, the old plan is refunded pro rata, but the switch is permanent.
| Legacy Cursor plan | Price | Comparable X Subscription plan |
|---|---|---|
| Pro | $20/month | Premium ($30) |
| Pro+ | $60/month | Super ($100) |
| Ultra | $200/month | Ultra ($200) |
| Teams Standard | $40/user | Premium seat ($30) |
| Teams Premium | $120/user | Super seat ($100) |
The deciding factor is which models you use. Legacy plans have a monthly Other Models pool that includes some third-party usage. On The X Subscription, every Claude or GPT request is billed on top. If most of your work runs on Claude, staying on legacy Pro is likely cheaper. If you mostly use Composer or Grok, or you already pay for SuperGrok or X Premium, the bundle can be better value.
Agent and Tab: day-to-day coding
Agent is the reason most people pay for Cursor. You describe a task, it searches the codebase, proposes or makes edits across files, runs commands such as your test suite, and shows you the diff to accept or reject. Plan mode writes a plan you can correct before any file changes, which is the cheapest way to catch a wrong approach on a large task. Ask mode answers questions without editing anything.
Project rules files let you write down your stack, test commands and conventions once, and the agent reads them on every request. As our AI coding assistants guide explains, that file is the single biggest lever on output quality. Good task scoping matters as much as the tool; our guide to writing better prompts applies directly.
Tab is Cursor’s autocomplete. It suggests multi-line edits and jumps to the next likely change, including in other files. Cursor’s legacy Pro plan advertised unlimited Tab; the new pricing pages do not say whether Tab is unlimited on The X Subscription, so check before you rely on it.
Cloud Agents, Projects and Bugbot
Cloud Agents are where Cursor pulls ahead of most editor assistants. Each one runs in an isolated cloud machine, so you can start several in parallel, close your laptop and come back to pull requests. Since September they can also run on your own machines or third-party sandboxes, and Projects (in beta, rolling out from 13 October) adds a coordinator that hands work to several agents with shared context. Cloud Agents and Automations start at Premium. Recent changelog entries add more: since 16 October Cloud Agents can build interactive canvases you share by link, a beta Environments API lets you create and manage their cloud environments in code, and an environment can now span up to 300 repositories. Since 6 October the Cursor iOS app can also see and reply to the agents running locally on your computer (Remote Control, on by default except for Enterprise organisations).
Bugbot reviews the diff of a pull request on each update, or on demand with cursor review or bugbot run. You can teach it rules with @cursor remember, and Bugbot Autofix starts a Cloud Agent to fix what it finds, billed at Cloud Agent rates. Bugbot has used usage-based billing since Cursor’s May 2026 pricing update and is not available on Plus. Security Review and Rollouts, added on 23 September, are for Teams and Enterprise only. If you host code on Cursor’s Origin, its app marketplace (opened to public submissions on 13 October) adds more than 20 integrations, including CodeRabbit and Greptile for review, Sentry and Linear.
Models: Composer, Grok, Claude, and OpenAI leaving
Cursor’s model list is wide, but the price you pay depends on which side of the pool a model sits. Per-million-token prices at the time of writing:
| Model | Input / output | Billed from |
|---|---|---|
| Composer 2.5 | $0.50 / $2.50 | Weekly pool |
| Grok 4.7 | $2 / $6 | Weekly pool |
| Claude Sonnet 5.5 | $2 / $10 | Prepaid credit |
| Claude Opus 5.5 | $4 / $20 | Prepaid credit |
| Gemini 4 Argon | $2 / $10 | Prepaid credit |
| GPT-5.6 Sol | $4 / $20 | Prepaid credit |
Our Claude review covers why many developers prefer Claude for agent work, and our Claude Sonnet 5.5 launch report covers that model.
OpenAI is leaving. On 28 August 2026 it said it will wind down the contract supplying its models to Cursor, with a proposed shutoff date of 12 November 2026, and that future models will not come to Cursor. Its help article says the final date will be confirmed later and lists two workarounds: your own OpenAI API key, which works only in local Chat and Agent (not Tab, Auto, Cloud Agents, Automations or the CLI), or the Codex extension inside Cursor. If GPT models are central to your workflow, our ChatGPT review covers OpenAI’s own coding tools.
Privacy and security
Privacy Mode is available on every plan. With it on, Cursor says your code is not used for training by Cursor or its model providers; with it off, Cursor may store and use prompts and code to improve its models (Cursor data use). It is on by default for Enterprise teams; the docs do not say clearly whether it is on by default for individual accounts, so check Settings after you sign up. Cloud Agents are the only feature that stores code, encrypted and temporarily. Cursor lists SOC 2, ISO 27001, ISO 42001 and AIUC-1 certifications. Whether The X Subscription shares data between Cursor, Grok and X accounts is not addressed in the docs at the time of writing.
What we don’t like
- Pricing churn. Cursor changed pricing in May, was acquired in August and was moved into The X Subscription in October.
- Opaque included usage. “4x Plus” tells you nothing about how many agent hours you get.
- Claude and GPT are extras. On the new plans, the models many developers prefer are billed from the first request.
- Bundling you may not want. Grok and X benefits are irrelevant to many developers, yet they shape the plans and share your usage pool.
Cursor vs GitHub Copilot and Claude Code
| Tool | Entry price (at the time of writing) | Where it runs | Best at |
|---|---|---|---|
| Cursor | Plus $8, Premium $30/month | Its own VS Code-based editor, CLI, cloud | Agent work across files, cloud agents, PR review |
| GitHub Copilot | Pro $10/month | Your existing IDE (VS Code, JetBrains and others) | Low-friction autocomplete and chat |
| Claude Code | Claude Pro $20/month ($17 billed annually) | Terminal and IDE extensions | Delegating whole tasks to Claude |
Copilot is the cheaper, lower-friction choice if you do not want to change editors. Claude Code suits terminal-first developers who want Claude without per-token billing on top of a subscription. Cursor is the strongest all-in-one editor if you will actually use Cloud Agents and Bugbot. More head-to-heads are in the Comparisons section.
Who should use Cursor, and who should look elsewhere
Use Cursor if you are a full-stack developer who hands multi-file tasks to an agent, a small team that wants Cloud Agents and automated PR review in one tool, or a Grok or X user who gets extra value from the bundle.
Look elsewhere if you are committed to JetBrains or Visual Studio and want the cheapest option (Copilot), you live in the terminal (Claude Code), you use Claude for nearly everything and dislike metered API costs, or you need GPT models after November.
Verdict
Cursor is still the most complete AI code editor for developers who delegate real work to agents, and its changelog shows frequent, substantial updates. The commercial wrapper is the weak point: pricing has changed three times this year, included usage is opaque, and third-party models now cost extra. Premium at $30 is the right plan for most developers; Super at $100 for people who run agents all day; Plus only to try the agent. Existing Pro subscribers who rely on Claude should stay where they are. More reviews are in the Reviews section.
Frequently asked questions
Is Cursor worth it in 2026?
For developers who delegate multi-file work to an AI agent, yes. Premium at $30 a month (at the time of writing) includes Agent, Cloud Agents, Bugbot and access to third-party models at API prices. If you mostly want autocomplete in your existing IDE, GitHub Copilot Pro at $10 a month is cheaper.
Is Cursor free?
There is a Free plan with limited Agent and Chat usage. Paid plans are sold as The X Subscription: Plus $8, Premium $30, Super $100 and Ultra $200 a month at the time of writing, with annual prices of $80, $300, $1,000 and $2,000.
What is the difference between Cursor Plus and Premium?
Plus ($8) gives you Agent with Cursor's own models (Grok and Composer 2.5) at fixed settings, with no Auto, Cloud Agents, Automations, Bugbot or third-party models. Premium ($30) has 4x the included usage of Plus and adds Cloud Agents, Automations, Bugbot, Origin code hosting and Claude, GPT and Gemini at API list price.
Do I have to switch from Cursor Pro to The X Subscription?
No. Existing Cursor plans continue unless you change or cancel them. If you switch, the old plan is refunded pro rata, but the switch is permanent. Legacy Pro maps to Premium.
Can I still use Claude and GPT models in Cursor?
Yes, on Premium and above, paid from prepaid credit at API list price (for example Claude Sonnet 5.5 at $2 / $10 per million input / output tokens). OpenAI has proposed ending its model supply to Cursor on November 12, 2026; the final date has not been confirmed.
