Cursor Review: A Powerful Editor With Pricing You Must Read Twice
Cursor is best understood as an editor first and an assistant second, because the AI is woven into editing, searching, reviewing and running code rather than sitting in a side panel. For developers who like steering an agent through a real codebase, that integration feels fast and natural. The trade-off is pricing complexity, especially after recent plan changes, so you should understand exactly which models are included before paying.
What it is
Cursor is an AI-first code editor developed by Anysphere, built on the open-source foundations of VS Code, which is why it feels immediately familiar to many developers. The difference is in emphasis: in a traditional editor, AI is an add-on, while in Cursor the agent is a central way of working. You can describe a feature, ask the agent to explore the codebase, and watch it propose coordinated changes across files, run commands and adjust its plan based on the results.
Key features
The agent is the headline feature. It can search file names, contents and project structure, apply edits across several files, run terminal commands such as tests or builds, and iterate when something fails. Tab completion complements this by predicting your next edit from recent changes, which can make routine editing feel noticeably quicker. Project rules and reusable instructions help the agent follow your conventions, while model context tools and hooks let teams connect extra tools and automate parts of a workflow. On higher plans, cloud agents and automations can run work in the background, and Bugbot can review pull requests for likely defects.
Pricing
Cursor now sells through a bundled subscription that also covers related chat and social features, and its current individual plans are Free at zero dollars, Plus at 8 dollars per month, Premium at 30 dollars, Super at 100 dollars and Ultra at 200 dollars, with Teams starting at 30 dollars per user per month. Usage is pooled and refreshes weekly, with larger pools on higher tiers. The detail that catches many buyers out is model inclusion: the Plus plan covers first-party models only, while third-party models become available on Premium and above but are billed separately at API list price, paid from prepaid credit.
What users like
Users commonly praise Cursor for its understanding of an existing codebase. Rather than answering in isolation, it can find relevant files, follow patterns already present in a project, and make edits that respect surrounding code. Many people also like that the editor itself feels polished and familiar, so the learning curve is mostly about learning how to prompt and review the agent, not learning a new editing environment. Debugging and feature work are frequent bright spots, with users reporting that multi-file changes and first drafts arrive much faster than by hand.
What users complain about
Complaints cluster around cost and predictability. Heavy users report that usage limits can feel restrictive, that it is hard to forecast how much agent work a plan will cover, and that plan changes have reduced the value they expected at the same price. Some users have also reported confusing billing or slow support when a plan change or an unexpected charge needed sorting out. On the product side, the agent can sometimes alter code that was working fine, or suggest a change that looks reasonable but misses an edge case, which means every diff still needs attention. These issues do not erase the productivity gains, but they do mean Cursor rewards users who set spending caps, watch usage and review changes closely.
Limitations to know
An agent that can edit many files and run commands is powerful, and that power is also the risk. It can introduce subtle bugs, weaken security checks, or drift away from the intended design if instructions are vague. Generated code may also rely on outdated library behaviour or on patterns that work in a demo but fail under real data. Treat Cursor output as a draft from a very fast junior colleague: useful, often impressive, but not ready to ship without reading, testing and, for sensitive areas, a second human review. It is also worth remembering that agent quality depends on the model selected and on how clearly you describe the task, so results will vary from session to session.
Who it is best for
Cursor suits developers who spend most of their day inside an editor and want agent help deeply integrated into that workflow. It is a strong fit for building features, refactoring across files, and working through a codebase you already know well. Teams that want shared rules, review tooling and central administration can use the team plans, while solo developers should weigh the Premium tier against the cheaper Plus tier, remembering that Plus covers first-party models only. It is less suited to someone who only wants occasional completions in an editor they already love, or to a buyer who wants a simple, all-inclusive price for heavy use of third-party models.
Verdict
Cursor is one of the most capable AI editors available, and its agent-led workflow can genuinely change how fast you move through a codebase. The reason it does not score higher here is pricing clarity: the headline plans are easy to read, but the real cost depends on which models you use and how much agent work you run. If you choose a plan with your model habits in mind, set a spending cap, and review agent output carefully, Cursor is an excellent tool. Our editorial score: 8.4/10 — our own judgement, not an aggregate of user ratings.