Cursor

Cursor is an AI-powered code editor built to make you extraordinarily productive. It combines the familiarity of VS Code with cutting-edge AI capabilities, offering intelligent code completion, natural language editing, and frontier AI models built right into your development environment.

Overview

Launched in 2023, Cursor has quickly become the go-to AI code editor for developers worldwide. It's designed to feel familiar while providing AI superpowers that transform how you write code. The editor is trusted by engineers at leading companies like Samsung, Stripe, Rippling, Perplexity, Ramp, Shopify, and many others.

Ownership has changed. SpaceX announced an agreement to acquire Anysphere on June 16, 2026 in a $60 billion all-stock transaction — widely described as the largest acquisition of a venture-backed startup on record — and closed it in mid-August 2026, folding Anysphere into its SpaceXAI unit alongside xAI. Cursor kept its name, its plans and its multi-vendor model list through the close. The thing worth noticing is structural rather than cosmetic: the company that builds Grok now also owns the editor that co-trains it, ships it on every paid plan, and publishes the CursorBench numbers it is measured on.

Cursor distinguishes itself by being purpose-built for AI coding, offering seamless integration of AI capabilities directly into the editor experience. Unlike traditional code editors with AI extensions, Cursor is designed from the ground up with AI as a core feature.

Key Features

• Tab Completion: Intelligent code completion that predicts your next edit.
• Parallel Agents: Run multiple agents at once on different parts of your codebase, locally or as cloud agents in isolated VMs.
• Codebase Awareness: Get answers from your codebase with secure codebase indexing and semantic search.
• Natural Language Editing: Write code using instructions and update entire modules with simple prompts (vibe coding approach).
• Frontier Intelligence: Cursor's own Composer 2.5 and the co-trained Grok 4.6 are included on paid plans; Claude Fable 5.1, Claude Opus 5, Claude Sonnet 5, the GPT-5.6 tiers and Gemini 3.8 Flash need Pro or higher.
• Agentic Tooling: Run tests, execute terminal commands, and use MCPs, skills, and hooks.
• Bugbot: Agentic code reviews on pull requests.
• Familiar Interface: Full VS Code compatibility with one-click migration.

How It Works

Cursor uses advanced AI models to understand your codebase and provide intelligent assistance. The process involves:

1. Codebase Analysis: Analyzes your entire project structure and context
2. Pattern Recognition: Identifies coding patterns and project-specific conventions
3. Intelligent Prediction: Predicts your next edit based on context and patterns
4. Natural Language Processing: Understands instructions in plain English
5. Real-time Assistance: Provides suggestions and completions as you type

Technical Architecture:

• Models (from cursor.com/docs/models): Cursor-provider models are Composer 2.5 and Grok 4.6, each in a standard and a Fast variant, with Grok 4.5 still listed. Everything else needs Pro or higher: Claude Fable 5.1, Claude Opus 5 and Claude Sonnet 5 plus a long tail of older Claude releases; GPT-5.6 Sol, Terra and Luna and the GPT-5 line; Gemini 3.8 Flash and earlier Gemini releases; and third-party entries including Kimi K3 and GLM 5.2.
• Agentic Engine: Orchestrates parallel agents for multi-file changes, including cloud agents in isolated VMs.
• Privacy: Privacy Mode guarantees code data is not used for training by Cursor or its model providers.
• Performance: Optimized for speed and responsiveness
• Extensibility: Full VS Code extension compatibility; MCPs, skills, hooks, and a marketplace
• Cross-platform: Available on macOS, Windows, and Linux, plus a CLI and an iOS app

Three things about that model list are worth spelling out.

OpenAI's current flagship is not on it. GPT-6 Astra shipped on September 4, 2026 and does not appear in Cursor's model picker; the newest OpenAI models Cursor offers are the GPT-5.6 tiers. If Astra is the model you want driving your repository, Codex has it and Cursor does not.

Claude Fable 5.1 is the one to reach for on long sessions. Its sticker price is unchanged from Fable 5 at $10 / $50 per million tokens, but a prompt-cache read now costs $0.25 per million instead of $1.00 — a 75% cut, and cache reads are most of the bill in an editor agent that replays a large cached prefix on every turn. Anthropic puts the saving at roughly 25% on typical use and up to 45% on highly agentic work.

Composer 2.5 is Cursor's own model, but not Cursor's own pretraining. Released May 18, 2026, it is post-trained from Moonshot's open-source Kimi K2.5 checkpoint. Moonshot retired the kimi-k2.5 and moonshot-v1 endpoints from its own API on August 31, 2026; that does not affect Composer, which runs on weights Cursor serves itself, but the base checkpoint can no longer be called at Moonshot.

Use Cases

Software Development

• Rapid Prototyping: Quickly build and iterate on applications
• Code Refactoring: Update entire classes or functions with natural language
• Feature Development: Accelerate feature implementation with AI assistance
• Bug Fixing: Get intelligent suggestions for debugging and fixes
• Documentation: Generate and update code documentation

Learning & Education

• Code Examples: Learn new patterns through AI-generated examples
• Best Practices: Discover industry-standard coding approaches
• Language Learning: Explore new programming languages with guided assistance
• Code Explanation: Understand complex code through AI explanations

Productivity Enhancement

• Boilerplate Generation: Create common patterns and structures quickly
• API Integration: Generate code for connecting to external services
• Configuration Management: Create and update config files
• Test Generation: Write comprehensive tests with AI assistance

Specialized Development

• Web Development: Generate HTML, CSS, JavaScript, and framework code
• Data Science: Create data processing and analysis scripts
• Mobile Development: Generate code for cross-platform applications
• DevOps: Create deployment scripts and infrastructure code

Pricing & Access

Prices in USD, from cursor.com/pricing. Every plan includes a set amount of model usage; on-demand usage beyond that is billed in arrears.

Hobby — Free

• No credit card required
• Limited Agent requests
• Limited Tab completions

Individual — from $20/month (Pro; Pro+ and Ultra tiers available)

• Everything in Hobby, plus:
• Extended limits on Agent
• Access to frontier models
• MCPs, skills, and hooks
• Cloud agents
• Bugbot on usage-based billing

Cursor recommends Pro+ for daily agent users and Ultra for agent power users.

Teams — $40/user/month (Standard; Premium tier available)

• Everything on Individual, plus:
• Centralized team billing and administration
• Team marketplace for internal rules, skills, and plugins
• Agentic code reviews with Bugbot
• Cloud agents and automations with shared team context
• Usage analytics and team-wide privacy mode
• SAML/OIDC SSO

Enterprise — Custom

• Everything on Teams, plus pooled usage, invoice/PO billing, SCIM seat management, repository/model/MCP access controls, audit logs, and priority support

Getting Started

Step 1: Download and Install

1. Visit cursor.com
2. Download Cursor for your operating system
3. Install the application
4. Launch Cursor

Step 2: Import Your Settings

1. Import your VS Code extensions, themes, and keybindings
2. Configure your preferred settings
3. Set up your development environment

Step 3: Start Coding

1. Open a project or create a new one
2. Start typing code and see AI suggestions
3. Use Tab to accept completions
4. Use Cmd+K (Ctrl+K on Windows/Linux) for natural language editing
5. Use Cmd+L (Ctrl+L on Windows/Linux) for AI chat

Step 4: Customize AI Behavior

• Model Selection: Choose between different AI models
• Privacy Settings: Configure Privacy Mode if needed
• Completion Settings: Adjust when and how suggestions appear
• Chat Preferences: Customize AI chat behavior

Best Practices

• Write Clear Comments: Use descriptive comments to get better suggestions
• Provide Context: Include relevant imports and function signatures
• Review Suggestions: Always review AI-generated code before accepting
• Use Natural Language: Leverage natural language editing for complex changes
• Iterative Development: Use chat for step-by-step problem solving
• Privacy Awareness: Enable Privacy Mode for sensitive projects
• Model Selection: Choose the right model for your specific task

Limitations

• Internet Dependency: Requires internet connection for AI features
• Learning Curve: Takes time to learn effective prompting techniques
• Cost: Subscription required for full features
• Privacy Concerns: Code may be sent to servers (unless Privacy Mode enabled)
• Accuracy: AI suggestions may not always be correct or optimal
• Context Limits: May not always understand complex project architectures
• Performance: May slow down on very large codebases

Alternatives

• Windsurf - The first agentic IDE, featuring Flow state tracking
• Claude Code - Anthropic's native CLI agentic coding tool
• Cline - Leading open-source autonomous agent for VS Code
• Aider - Premier CLI-based AI pair programmer
• GitHub Copilot - AI coding assistant with IDE integration
• Replit AI - AI coding assistant in online IDE

Community & Support

• Documentation: docs.cursor.com
• Community Forum: forum.cursor.com for discussions and support
• Discord: Community server for real-time help
• Twitter: @cursor_ai for updates
• GitHub: Open-source components and examples

What Developers Say

"Cursor is at least a 2x improvement over Copilot. It's amazing having an AI pair programmer, and is an incredible accelerator for me and my team." - Ben Bernard, Instacart

"The Cursor tab completion while coding is occasionally so magic it defies reality - about ~25% of the time it is anticipating exactly what I want to do." - Kevin Whinnery, OpenAI

"Cursor is hands down my biggest workflow improvement in years" - Sawyer Hood, Figma

"Cursor is the best AI developer tool right now, avoid it at your own peril" - Logan Kilpatrick, Google