Cursor AI: write software by describing what you want
Cursor AI is a code editor built around language models. It indexes your repository, answers questions about it, and turns a short instruction into real edits across many files — while you stay in control of every change.
// prompt: add retry with backoff to fetchUser() async function fetchUser(id, tries = 3) { for (let i = 0; i < tries; i++) { try { return await api.get(`/users/${id}`); } catch (e) { await sleep(200 * 2 ** i); } } throw new Error("fetchUser failed"); }
An editor where the model knows your code, not just the open file
Most autocomplete tools only see a few lines around the caret. Cursor AI builds a semantic index of the entire project, so suggestions, chat answers and agent edits are grounded in the functions, types and conventions you already use.
Predictive editing
The Tab model guesses your next edit, not only the next word. It can rewrite several lines at once and jump the cursor to where the change continues.
Ask the codebase
Pose a question like “where is the session token refreshed?” and get an answer with references to the exact files and symbols involved.
Multi-step tasks
Agent mode plans a change, edits files, runs terminal commands and fixes failing tests in a loop, showing a reviewable diff at the end.
Edit in place
Select a block, press Ctrl+K, describe the change. The result appears as a diff you accept or reject line by line.
Precise references
Pull files, folders, docs pages or git history into a prompt with the @ symbol to steer the model toward exactly the right material.
Project rules
Store team conventions in rule files so every generated snippet follows your naming, testing and architecture choices automatically.
Familiar surface, very different workflow
Cursor AI is based on the VS Code foundation, which means your keybindings, themes and most extensions carry over on the first launch. The learning curve is about prompting well, not about relearning an editor.
The difference shows up in how work gets split. Routine changes — renaming an API across modules, writing boilerplate tests, migrating a config format — become one instruction. Your attention moves to reviewing intent and edge cases.
Model choice is yours
Cursor AI lets you pick between several frontier language models for chat and agent work, and switch per task: a fast model for small edits, a stronger reasoning model for refactors.
Open a project
Cursor AI indexes the folder in the background and keeps the index fresh as you type.
Describe the goal
Write what should change in plain English inside chat or the agent panel.
Review the diff
Each touched file shows green and red lines you can accept, edit or discard.
Iterate quickly
Follow-up prompts refine the same change without losing context.
Everything you need to start with Cursor AI
Cursor download
Installers for Windows, macOS and Linux, system requirements and first-launch steps.
Read the guide →Cursor settings
Tune models, privacy, indexing, rules and keyboard shortcuts for your workflow.
Configure the editor →Cursor library
A curated collection of rules, prompts and reusable context for popular stacks.
Browse the library →Cursor AI questions, answered
Try Cursor AI on your own repository
Open a real project, ask one question about it, and judge the answer yourself. It takes less than five minutes.
Start with Cursor AI