✓ Reviewed by The Future Signal
✓ Reviewed by The Future Signal
Cursor is still one of the most complete AI coding tools available. Its agents can work in parallel, in the cloud and on long projects, and built-in bots now check code, security and deployments. It suits engineering teams of every size, especially those with strict security needs. Since SpaceX bought it, Cursor’s own and Grok models are cheaper, but Claude, GPT and Gemini cost extra and GPT access may end. We recommend it for software teams that check their model usage and new pricing before committing.
Cursor’s main impact is more output from the same engineering team. Routine work such as small bug fixes, tests and refactors can be handed to agents, while engineers spend their time reviewing and on harder problems. Cloud agents and Projects keep work moving outside office hours, which shortens delivery times. Automatic code review, security checks and production monitoring reduce the cost of mistakes reaching customers. For technical founders, Cursor can stand in for extra headcount at the prototype stage. The cost picture depends on model choice: teams using Cursor’s own models can lower their per-seat spend, while teams relying on Claude or GPT need to budget for usage on top.
Compare plans and pricing to find the best option for your needs.
Compare plans and pricing to find the best option for your needs.
Last Updated: 25 October 2026
Cursor is an AI coding tool that lets developers hand whole tasks to AI agents, then review the result. It is built for software teams of every size. Its biggest strengths are a mature agent workflow, strong security credentials and fast product development. Since our last review, SpaceX has bought Cursor and its plans have moved into “The X Subscription”. Cursor’s own and Grok models are now cheaper, but Claude, GPT and Gemini cost extra, and OpenAI plans to withdraw its models. Our verdict: still highly recommended for engineering teams, but check which AI models you rely on before you buy or switch.
Estimated reading time: 11 minutes
How much of your developers’ week goes on routine changes, small bug fixes and tests? Cursor is designed to take that work off their hands.
Cursor is an AI code editor and coding agent made by Anysphere, a San Francisco company founded in 2022 by four MIT students. It looks like Visual Studio Code, the editor most developers already know. The difference is that you can describe a change in plain English and an AI agent will find the right files, make the edits, run tests and show you the result for approval.
It became popular because it made AI feel like part of the developer’s normal workspace, not a separate chat window. Cursor says 64% of Fortune 500 companies now use it, though that is the company’s own figure.
Our previous review was published on 25 July 2026. Three changes since then matter for buyers:
| Date | Change | Why it matters |
|---|---|---|
| 14 Aug 2026 | SpaceX completed its acquisition of Cursor, reported at $60 billion | New owner, new model strategy, new terms |
| 29 Aug 2026 | OpenAI said Cursor’s direct access to its models would end on 12 November 2026 | GPT users may lose access inside Cursor |
| 12 Oct 2026 | New customers moved to “The X Subscription”, which bundles Cursor, Grok and X Premium | Prices, plan names and model billing all changed |
Cursor is built for people who write and maintain software. The clearest fits are:
Who should probably look elsewhere?
Since Cursor 3 in April 2026, the app is built around an Agents window. You can run several agents at once, each in its own copy of the code, and compare results from different models.
Business value: one engineer can supervise several pieces of work at the same time. A typical use is asking two agents to fix the same bug and keeping the better fix.
Cloud agents run on Cursor’s servers, or on machines your company manages. You can start them from the desktop app, the web, the iPhone app, Slack or Linear.
Business value: work continues when a laptop is closed. A bug report in Slack can become a ready-to-review pull request without anyone opening an editor.
Projects, in beta since September, uses a “coordinator” agent to plan larger jobs such as a new feature or a migration. It hands tasks to other agents, remembers context over weeks, and can run on a schedule.
Business value: big jobs stop restarting from zero each session, which cuts supervision time on multi-week work.
Origin is Cursor’s own place to store code, launched in August as an alternative to GitHub. It has built-in agents, two-way sync with GitHub and an app marketplace for deployment, testing and monitoring tools. It is included from the Premium plan.
Business value: small teams can keep code, agents and reviews in one place. Larger teams will likely keep GitHub for now and use the sync.
Bugbot reviews pull requests for bugs. Security Review, launched in September for Teams and Enterprise, checks every change for vulnerabilities and offers a one-click fix. Rollouts watches production after a release and can propose a revert if something breaks.
Business value: routine checks happen on every change, not only when a senior engineer has time.
Cursor includes its own Composer models and SpaceX’s Grok models in the plan. Claude, GPT and Gemini are available on Premium and above, billed separately. On Teams and Enterprise, Cursor Router can pick a model for each request to balance cost and quality.
Business value: you can match cost to the task, but only if someone keeps an eye on usage.
Developers who know Visual Studio Code will feel at home within minutes. Settings and extensions carry over, and the basic agent chat needs no training.
Mastering Cursor takes longer. Parallel agents, cloud environments, Projects, model choice and the new billing system add a lot of concepts. Documentation is extensive and the community forum is active.
Non-developers can build simple prototypes, but someone who can read code should still review the output.
Future Signal Tip: Start with one agent workflow, such as fixing small bugs from your issue tracker. Add parallel agents and Projects only once your team trusts the review process.
Cursor’s agents are fast and capable on everyday coding tasks, and its customer list suggests they hold up in large codebases. PayPal, for example, told Cursor that work that took four sprints now takes one. These are vendor-published testimonials, so treat them as examples, not guarantees.
Model quality now depends on which model you choose. On SpaceX’s own published benchmarks, Grok 4.7 is competitive with leading models on some coding tests but behind Claude Fable 5.1 on Cursor’s own benchmark and on terminal tasks. We found no independent study that measures Cursor as a whole product, so evidence on accuracy and reliability remains limited.
Cursor connects to the tools engineering teams already use:
There are official APIs for agents and environments. Cursor does not integrate directly with office tools such as Microsoft 365 or CRM systems, which is not its purpose.
This is where the biggest change has happened. New customers now choose from The X Subscription, which shares one weekly usage allowance across Cursor, the Grok apps and X.
| Plan | Price (USD/month) | What you get in Cursor |
|---|---|---|
| Free | $0 | Limited agent and chat use; no third-party models |
| Plus | $8 | Cursor with Grok and Composer models only |
| Premium | $30 | 4x Plus usage; cloud agents, Origin, AI code reviews; Claude, GPT and Gemini at API prices |
| Super | $100 | 15x Plus usage; priority access at busy times |
| Ultra | $200 | 50x Plus usage; earliest access to new features |
| Teams | From $30 per user | Premium, Super or Ultra seats with SSO, analytics and central billing |
| Enterprise | Custom | Pooled usage, audit logs, access controls, invoicing |
Prices from cursor.com and x.ai on 25 October 2026, excluding taxes. Annual prices are not shown on Cursor’s pricing page.
What does this mean in practice? If your team mostly uses Cursor’s and SpaceX’s models, you get more for less. SpaceX says each plan includes at least twice the coding usage of the comparable old Cursor plan, and Teams seats are $10 cheaper.
If your team mostly uses Claude or GPT, costs can rise. The old $20 Pro plan included $20 of third-party model use. Now that use is charged on top. The New Stack’s example: a developer who paid $20 and used $20 of Claude would now pay $30 plus $20, or $50.
Callout — before you switch plans: Existing Cursor customers can keep their current plan, price and features for now. Switching to the new plans is permanent. Check your usage by model in Cursor’s dashboard first.
Cursor has become a test case for a question every AI buyer will face: what happens when your tool’s owner also sells AI models?
Until this year, Cursor’s appeal was neutrality. It gave developers the best models from every lab inside one editor. Under SpaceX, the economics now favour Grok and Cursor’s own models, while rivals’ models cost extra or, in OpenAI’s case, may leave. The product itself has never been stronger, with agents that plan, code, review and monitor releases.
We expect Cursor to stay a market leader in AI coding, helped by SpaceX’s computing power and its head start on agent workflows. The open question is whether developers accept a more vertically integrated tool or move to model-neutral alternatives.
Businesses with large engineering teams should keep Cursor on the shortlist, but treat model access as a contract risk. Watch three things: the outcome of OpenAI’s November cutoff, any forced move off legacy plans, and independent benchmarks of Grok models against Claude and GPT.
Overall rating: 8.5 / 10
Cursor is still one of the best AI coding tools for engineering teams that want to delegate real work to AI agents. Its agent workflow, review bots and security credentials are hard to match.
Value for money now depends on your model choice. For teams happy with Cursor’s own and Grok models, it is better value than before. For Claude- or GPT-heavy teams, it may cost more than a dedicated alternative.
Future Signal recommends Cursor for software teams that review its new pricing against their actual usage first. It is not the default choice for every developer it was a few months ago.
Yes, for most engineering teams. The product has improved, and plans now start lower. The value depends on which models you use, so check your usage before you choose a plan.
Yes, on Premium ($30) and above. Claude is billed separately at Anthropic’s API prices from prepaid credit. Anthropic has said it will keep supporting Claude in Cursor.
OpenAI has said Cursor’s direct access to its models will end on 12 November 2026. Cursor said it was in talks with OpenAI. Teams that depend on GPT should plan an alternative.
Not yet. Existing customers can keep their current plan, price and features. Switching is permanent, and no date has been given for moving remaining customers.
Not when Privacy Mode is on. Cursor says it will not use customer data for training in Privacy Mode and has zero-retention agreements with model providers. Admins can enforce it for the whole team.
Teams at $30 per user is the natural starting point, because it adds SSO, central billing and usage reports. Set spending caps for third-party models from day one.
Only for simple prototypes. Someone who can read code should review anything that goes into production.
Cursor’s biggest strengths are its mature agent workflow, its coverage of the whole delivery cycle from coding to monitoring, and strong enterprise security. Its biggest limitations are new: extra charges for Claude, GPT and Gemini, uncertain GPT access, and vendor questions after the SpaceX acquisition.
The ideal customer is a software team that wants to delegate routine and long-running engineering work and is comfortable with Cursor’s own and Grok models doing most of it.
Practical advice: if you already use Cursor, stay on your current plan until you have checked your usage by model. If you are new, start a Teams pilot with spending caps, enforce Privacy Mode, and compare the monthly cost with one alternative before you roll it out.
Our final assessment after evaluating features, performance, value, and business impact:
Cursor deserves a place on the shortlist of any business with an engineering team. Its agent workflow is mature, it now covers review, security and deployment as well as coding, and its security credentials meet most enterprise requirements. Teams that are happy to use Cursor’s own and Grok models will find it better value than before, with Teams seats now $30 per user.
It is a harder call for teams built around Claude or GPT. Those models now cost extra on every new plan, and OpenAI plans to withdraw its models in November. Those teams should compare Claude Code, OpenAI Codex or GitHub Copilot before committing. Larger organisations should also review the change of ownership and subscription terms with procurement.
Overall, Cursor is worth paying for, provided you choose a plan based on the models your team actually uses.
We may earn a commission if you purchase through our links, at no extra cost to you. This helps support independent reviews.
Explore this tool compared with other highly rated AI tools…
Explore other highly rated AI tools we’ve reviewed in this category.
