Coding · Anysphere
Cursor
An AI-first code editor: a full editor where an AI agent reads your whole project, writes and changes code across files, and runs tasks, with you reviewing every change.
Visit Cursor →What it does
Agent
Describe a change in plain language and the agent plans it, edits across multiple files, runs commands and shows you a diff to accept or reject, using frontier models or Cursor's own Composer model.
For work Building features, refactoring and fixing bugs faster, while you stay the reviewer.
Tab autocomplete
Predicts your next edit, not just the next word, and suggests multi-line changes as you type. Unlimited on the paid plans.
For work The moment-to-moment speed-up of actually writing code.
Codebase understanding
Indexes your project so you can ask questions about how it works and the agent can make changes that fit the code that is already there.
For work Getting oriented in an unfamiliar codebase and making consistent changes across it.
Bugbot and cloud agents
Bugbot reviews changes for likely bugs, and cloud agents run longer tasks in the background rather than on your own machine.
For work Catching mistakes before they ship and offloading longer jobs.
Best for
People who write code, from professional developers to technical founders, who want an editor where an AI agent does the heavy lifting under their review.
Not for
People who do not code: it is a developer's editor, not a no-code app builder, and its output still needs someone who can read and judge the code.
Real work with it
A developer
Add a feature to an existing project
- Open the project so Cursor can index it, and describe the feature in plain language.
- Let the agent draft the changes across the files it needs, then read the diff.
- Accept the parts that are right, correct or reject the rest, and run the tests.
Human checkpoint Read every changed line and run the tests before you commit: an agent's confident diff can still be subtly wrong or insecure.
A developer new to a codebase
Understand and fix a bug in unfamiliar code
- Ask Cursor how the relevant part of the system works and where the bug likely is.
- Have the agent propose a fix, and ask it to explain why.
- Apply the fix, reproduce the original bug, and confirm it is gone.
Human checkpoint Make sure you understand the fix yourself. Shipping code you cannot explain is how a small bug becomes an incident.
The part that stays yours
The agent can write a lot of code quickly, but you own what ships. Reading every diff, understanding what changed, testing it, and refusing changes you cannot vouch for are the developer's job. Speed without review is how bugs and security holes get in.
Plans and pricing
Hobby
US$0
Trying Cursor: limited agent use, no card needed.
Pro
US$20/mo
Individual developers: unlimited Tab, all models, cloud agents and included usage.
Ultra
US$200/mo
Power users: about twenty times Pro's model usage.
Teams
US$40/user/mo
Teams needing shared context, SSO and central billing.
Free Hobby tier to try it. Pro is US$20/month and includes about US$20 of model usage at API prices before usage-based billing; a Pro+ tier sits above it, and Ultra is US$200/month with roughly twenty times Pro's usage. Team plans start at US$40 per user per month, with custom Enterprise.
When to pick something else
GitHub Copilot
You want AI assistance inside the editor you already use, such as VS Code or a JetBrains IDE, rather than switching editors.
Claude Code or the Gemini CLI
You prefer an agent in the terminal that works across your project without a graphical editor.
A hosted assistant's chat
You only occasionally need a snippet or an explanation and do not want a full coding tool.
Watch-outs
- The agent writes code confidently whether or not it is correct. Unreviewed changes can introduce bugs, security holes or subtle behaviour changes, so read every diff.
- Paid plans bundle a set amount of model usage and then bill by usage: heavy agent use can run up cost beyond the headline price, so watch your usage.
- It is a developer tool. Without the ability to read code, you cannot judge or safely use what it produces.
- Sending a private codebase to an AI editor is a data decision: check your employer's policy and the plan's data handling before you point it at sensitive code.
- Tiers, limits and the underlying models change often, so re-check what your plan includes before you rely on it.