Cursor

Cursor is an AI-first code editor built on VS Code’s foundation, redesigned so that AI assistance is integrated into every layer of the editing experience rather than added as a plugin. Where most editors treat AI as an autocomplete layer, Cursor ships with multi-file editing, an autonomous agent mode, full codebase indexing, and access to multiple frontier language models from a single subscription. It is used by individual developers and engineering teams who want AI to handle the mechanical parts of coding — boilerplate, refactors, debugging, and multi-step implementation tasks — while they stay focused on higher-level decisions.

How Cursor works

Cursor indexes your entire project directory so the AI understands how files relate to each other, not just the file you have open. This codebase context is what separates it from tab-completion tools that only see the current file.

The main interaction modes are:

Tab autocomplete predicts multi-line edits based on what you are about to type. Unlike single-token completion, it reads the surrounding code and suggests larger coherent changes, including deletions and rewrites, that you accept with a single keystroke.

Chat and Cmd+K let you ask questions about your code or give targeted instructions. You can highlight a block and ask the AI to explain, refactor, or rewrite it, or ask open questions about how a function behaves without leaving the editor.

Composer is the multi-file editing mode. You describe a change in natural language — “add rate limiting to all API routes” or “extract this logic into a shared utility and update all callers” — and Composer proposes edits across whichever files are affected. You review the diff and apply or reject changes before they land.

Agent mode goes further. You give it a goal and it plans its own steps: it reads files, writes code, runs terminal commands, checks output, and loops until the task is done or it needs your input. Agent mode suits tasks like scaffolding a new feature, fixing a failing test suite, or setting up a new service from scratch.

Cursor supports MCP (Model Context Protocol), the open standard for connecting AI tools to external data sources. This lets you hook the editor into databases, APIs, documentation systems, or internal tools so the AI can query live data as it works.

Who uses Cursor?

Cursor is used primarily by software engineers who write code professionally every day, from solo developers and freelancers to engineering teams at startups and larger companies. It suits developers who are comfortable with a code editor environment and want AI to accelerate their existing workflow rather than replace it. Teams adopting Cursor typically use it to cut time on repetitive implementation work — boilerplate, tests, migrations, and multi-file refactors — while keeping humans in the loop on architecture and review.

What you can build or do with Cursor

Multi-file refactors at scale: Use Composer to rename a data model and propagate the change across every file that references it, reviewing the full diff before committing anything.

Feature implementation from a spec: Describe a feature in plain language and let Agent mode scaffold the code, wire up tests, and fix type errors across the codebase, checking in with you only when it hits a genuine ambiguity.

Codebase Q&A: Ask questions like “where is user authentication handled?” or “what does this function return when the input is empty?” and get answers grounded in your actual code rather than generic documentation.

Test generation: Ask Cursor to write unit tests for a module it can read in context, covering edge cases you specify or ones it infers from the implementation.

PR review without context-switching: Pull up a PR inside the editor, ask the AI to explain what changed and flag anything worth scrutinising, without opening GitHub in a browser.

Greenfield project setup: Start a new project by describing the architecture you want — framework, folder structure, tooling — and let Agent mode generate the scaffold and initial configuration files.

Is Cursor free?

Cursor has a free tier with limited AI completions and access to slower model variants — enough to try the editor and evaluate whether it fits your workflow. The Pro plan at $20 per month gives a credit-based monthly allowance across frontier models including Claude, Gemini, and GPT-4o. Business at $40 per user per month adds centralised admin controls, SSO, and privacy mode (which disables model provider data retention). An Ultra tier at $200 per month targets developers who want maximum daily usage and priority access to new features.

The credit-based billing model means heavier models like o1 or Claude Opus cost more per request than lighter ones. Annual billing on Pro reduces the effective monthly cost to around $16.

How to get started with Cursor

Download Cursor from cursor.com — it installs as a standalone app and, on first launch, imports your VS Code settings, extensions, and keybindings automatically, so the transition is immediate. Connect your account, choose a plan or start the free tier, and open an existing project directory. Cursor indexes the project in the background. From there, Tab starts working immediately, and you can open the Composer panel (Cmd+I on Mac) to try a multi-file instruction on something low-stakes to see how it reasons about your codebase.