Overview
What does Cursor do?
Cursor is a fork of VS Code rebuilt so that an AI agent is a first-class part of the editor rather than a sidebar bolted on. It indexes your codebase, so questions and instructions are answered against your actual project rather than a generic model prior.
Day to day it operates on a spectrum. At the light end it is very fast multi-line autocomplete that predicts your next edit, including the next place you were going to jump to. At the heavy end, Agent mode takes a written instruction, edits files across the project, runs terminal commands and loops on errors.
Because it is still an editor, the human stays in the loop by default: every change surfaces as a reviewable diff you accept or reject. That makes it the gentlest on-ramp of the serious coding agents.
Features
Cursor features
Codebase indexing
Embeds your repository so @-mentions and questions resolve against real files, not guesses.
Tab completion that predicts edits
Its standout feature — suggests not just the next characters but the next edit and the next cursor location.
Agent mode
Give it a task and it plans, edits multiple files, runs commands and iterates on failures inside the editor.
Background and cloud agents
Kick off longer tasks that run remotely while you keep working, then review the resulting branch.
Rules files
Per-project `.cursor/rules` encode conventions so output matches house style without repeating yourself.
Model choice
Route between frontier models from several providers, or bring your own API key.
Integrations
Integrations and how to connect Cursor
What Cursor connects to, and the order to set it up in.
Install and import
Download Cursor and import your VS Code extensions, themes and keybindings in one click — the migration is deliberately near-frictionless.
Open a project and let it index
The first index pass takes a few minutes on a large repo. Quality of answers improves noticeably once it completes.
Add rules
Create `.cursor/rules` describing conventions, preferred libraries and testing approach. This is the highest-leverage configuration step.
Connect MCP servers
Add MCP servers in settings to give the agent access to your database, ticket tracker or docs during a task.
Enable background agents
Connect GitHub so longer tasks can run in the cloud and come back as a branch you review.
Fit
Who is Cursor for?
Developers already living in VS Code
The best fit of any tool here. The migration cost is close to zero and the productivity gain is immediate.
Teams that want a human reviewing every diff
The accept/reject flow is the safest default among agentic coding tools.
Engineers doing long unattended automation
Workable via background agents, but a terminal-native agent is a better fit for scripted, headless runs.
Non-developers
Poor fit. It is an IDE. The vocabulary and workflow assume you write software for a living.
Use cases
Best use cases for Cursor
Feature work in a familiar codebase
Where you know roughly what you want and reviewing diffs quickly beats writing them by hand.
Fast iterative prototyping
The tab-completion loop makes exploratory coding dramatically faster than typing.
Learning an unfamiliar area of a repo
@-mention files and ask questions grounded in the real code.
Incremental test writing
Generate tests file by file and accept them individually, which keeps quality controllable.
Pricing
Cursor pricing
Per-seat subscription with included usage, plus overage. Free tier: Yes — limited agent requests and completions.
Hobby
Free
Enough to evaluate the editor properly.
- Limited agent requests
- Limited tab completions
- Two-week Pro trial
Pro
Most common$20/user/month
The standard individual plan.
- Extended agent limits
- Unlimited tab completions
- Background agents
Ultra
$200/user/month
For engineers running agents continuously.
- Roughly 20x Pro usage
- Priority model access
- Highest limits on long tasks
Teams
$40/user/month
Org controls on top of Pro.
- Centralised billing
- SSO and admin dashboard
- Privacy mode enforced org-wide
Pricing last verified . Vendors change pricing frequently — always confirm on Anysphere’s own pricing page before buying.
User sentiment
What other users say about Cursor
A summary of publicly available feedback about Cursor — not paid testimonials.
Cursor converts trials to paid seats at a rate few developer tools match, and the reason users give is almost always the tab-completion model rather than the agent. Criticism concentrates on pricing changes and on usage limits being hard to reason about mid-month.
What users praise
- Next-edit prediction is repeatedly described as the feature people cannot give up
- Zero-friction migration from VS Code
- Diff-review flow keeps humans genuinely in control
- Codebase indexing makes answers specific rather than generic
Common complaints
- Pricing and limit structures have changed more than once, frustrating long-term subscribers
- Usage can be consumed faster than expected on agent-heavy days
- Agent mode is less autonomous than terminal-native competitors on long tasks
- Occasional lag and memory use on very large repositories
We publish individual user reviews only once they have been collected and attributed to a real source. None have been collected for Cursor yet. Used it? Send us your review.
Sentiment summarised from: Cursor — official documentation, Cursor — pricing.
Verdict
Is Cursor worth it?
Pros
- Fastest path from a normal editor to an agentic workflow
- Best-in-class inline completion
- Human review is built into the loop by default
- Generous enough free tier to evaluate honestly
Cons
- Pricing model has shifted repeatedly
- Less capable than terminal agents on long unattended tasks
- Heavy on system resources in big projects
- Integration surface is narrower than MCP-first competitors
Assessed from vendor documentation, pricing pages and public user sentiment — not a hands-on trial. How we assess agents.
FAQ
Cursor frequently asked questions
Is Cursor just VS Code with a plugin?
No. It is a fork of VS Code, which is why extensions and keybindings carry over, but the indexing, completion model and agent are built into the editor rather than added through the extension API.
Does my code get sent anywhere?
By default requests go to model providers. Privacy Mode stops code being retained for training, and Teams plans can enforce it across the organisation.
Can Cursor work offline?
No. The agent and completions require a network connection to a model provider.
Cursor or Claude Code?
Cursor if you want an interactive editor with a human reviewing each diff. Claude Code if you want long autonomous runs, CI automation and terminal scripting. Many teams pay for both.
Compare