What is Cursor? An honest review of the AI code editor

I've used Cursor as my main editor for the better part of a year. This is what it is, what it does well, where it quietly drains your wallet, and whether it deserves the hype.

What is Cursor, exactly?

Cursor is a standalone desktop app built on top of VS Code. Anysphere, the company behind it, forked Microsoft's open-source editor and rebuilt the AI layer from the ground up. That decision matters. It means your existing VS Code extensions, themes, and keybindings mostly carry over, so switching costs almost nothing.

What makes it more than "VS Code with a chatbot" is how deep the AI reaches. Cursor reads your whole codebase, understands the relationships between files, and can make coordinated edits across many of them at once. It's an AI code editor in the real sense, not a plugin sitting on the side.

If you've heard people talk about "cursor ai" and weren't sure whether that's a separate product, it isn't. Cursor and Cursor AI are the same thing. The AI is the point.

What does Cursor do?

Three features carry the product, and they're worth separating because people conflate them.

Tab (autocomplete). As you type, Cursor predicts your next edit, not just the next word. It'll suggest a multi-line change, jump you to the next spot that needs updating, and chain edits together. This is the feature that hooks people. After a week you feel slow in any other editor.

Chat. A side panel where you ask questions about your code, paste errors, or request changes. It sees the files you reference and answers with your actual code in mind, not generic Stack Overflow boilerplate.

Agent. The one that changed everything. You describe a task in plain English, and Cursor plans it, edits multiple files, runs terminal commands, and checks its own work. This is where "what does cursor do" stops being about autocomplete and starts being about delegation.

Since the 2.0 release in late 2025, the whole interface reorganized around that agent. The default view is now centered on outcomes and running agents rather than the file tree, though the classic IDE view is one click away when you want to read the code yourself.

The 2.0 shift: Composer and parallel agents

Cursor 2.0, which landed in October 2025, was the biggest change since launch, and it's the version most people are actually reviewing now.

The headline is Composer, Cursor's own coding model. Anysphere claims it runs about four times faster than similarly capable models and finishes most turns in under 30 seconds. In practice that speed is the real feature. A fast, decent model you'll actually use beats a brilliant one you wait 90 seconds for. Composer was trained specifically to navigate large codebases with tools like semantic search, and it shows on bigger projects.

The other big move is running multiple agents in parallel. Using git worktrees, Cursor can spin up several agents that don't step on each other, and you can even have a few of them attempt the same problem so you pick the best result. For thorny tasks this genuinely improves the output. There's also a native browser testing tool now, so the agent can load your app, check its own changes, and iterate until things work.

You're not locked into Composer, to be clear. You can still point Cursor at frontier models from other providers, including Claude and GPT, and pick per task.

Cursor pricing in 2026

Here's where you need to pay attention, because Cursor moved to usage-based billing and the plans read simpler than they bill.

The free Hobby plan gives you limited agent requests and Tab completions, no credit card required. It's a real trial, and fine for light use.

Pro is $20 a month. You get unlimited Tab completions, extended agent limits, access to frontier models, and a pool of usage credit for premium requests. This is the plan most individual developers should start on.

Above that, Pro+ at $60 a month gives roughly three times the usage, and Ultra at $200 a month gives around twenty times the Pro usage plus early access to new features. Teams is $40 per user per month with SSO, shared rules, agentic code reviews, and centralized billing. Enterprise is custom with pooled usage.

The catch worth stating plainly: since mid-2025, your subscription is effectively a pool of model usage, not a fixed number of requests. Simple autocompletes are cheap. But heavy work, premium models, Max mode with big context windows, and large-codebase runs burn credits faster than you'd expect. It's entirely possible to blow through a Pro pool in a busy week and get metered. Check the usage dashboard, or you'll be surprised.

Where Cursor genuinely shines — and where it falls short

Cursor vs the alternatives

GitHub Copilot is the obvious comparison. Copilot is cheaper, more predictable, and deeply tied to the GitHub ecosystem, but its agent and codebase awareness still trail Cursor's. If you want the strongest AI coding experience and will pay for it, Cursor wins. If you want good-enough autocomplete at a flat, predictable price, Copilot is the safer pick.

Windsurf and the newer wave of AI editors are closing in, and VS Code's own AI features keep improving. But as of mid-2026, Cursor is still the tool other tools get measured against.

• Best programming AI roundup — six tools ranked by how you actually code
• Best code review tools — pair with your editor stack

Who should actually use Cursor?

Use Cursor if you write code most days, work in a real codebase rather than one-off scripts, and want AI that acts, not just suggests. For professional developers, it's the current default, and Pro at $20 is an easy yes.

Skip it, or stay on the free tier, if you code occasionally, are learning the basics and want to understand what you write, or simply can't stomach usage-based billing. Copilot or the Hobby plan will serve you better.

The honest summary is that Cursor delivers on most of its promise. It's the rare AI tool where the productivity gain is real and repeatable, as long as you go in knowing you still have to read the code and watch the meter.

FAQ

What is Cursor in simple terms?

Cursor is an AI code editor — a version of VS Code with a built-in AI that can autocomplete, answer questions about your code, and edit multiple files on its own from a plain-English request.

Is Cursor AI free?

There's a free Hobby plan with limited agent and Tab usage. Serious daily use realistically needs the $20 Pro plan, and heavier use can cost more under the usage-based system.

What does Cursor do that VS Code doesn't?

It reads your entire codebase and makes coordinated edits across many files, runs an agent that can plan and execute tasks, and offers a much stronger autocomplete than stock VS Code with a basic extension.

Is Cursor better than GitHub Copilot?

For agent capability and codebase awareness, yes, Cursor leads in 2026 — see our full Cursor vs Copilot comparison. For predictable pricing and tight GitHub integration, Copilot still has an edge. It depends on what you value.

Do I need to know how to code to use Cursor?

It helps a lot. Cursor speeds up people who already code and can review its output. If you can't tell when the AI is wrong, you'll ship its mistakes.

What is Composer in Cursor?

Composer is Cursor's own fast coding model, introduced in version 2.0. It's built for speed and for navigating large codebases, and you can still switch to other frontier models per task.

Final verdict

Cursor is the AI code editor we'd recommend first for anyone who codes daily. Pro at $20/month pays for itself in Tab alone; the agent is now good enough for real delegation, not toy snippets. Just read what it writes and keep an eye on usage if you live on premium models.