Download Cursor

About Cursor — The AI Code Editor Built to Replace Your IDE

Cursor was founded in 2022 by Anysphere with a single thesis: the code editor is the last major developer tool that has not been rebuilt around AI. While other companies bolted AI features onto existing editors through extensions and plugins, Anysphere forked VS Code and rebuilt the editing experience from the ground up — making AI a first-class citizen at every layer of the IDE.

Today, Cursor is used by over one million developers worldwide. The editor ships Tab completions that predict multi-line edits, Composer for simultaneous multi-file changes, agent mode for autonomous coding, and @codebase for project-wide context-aware answers. Every VS Code extension, theme, and keybinding transfers in one click.

Cursor History and Mission — April 2026

  • Founded in 2022 by Anysphere, a San Francisco AI company dedicated to building the IDE of the future
  • Forked from VS Code's open-source codebase (Code OSS) — all extensions, themes, and keybindings transfer natively
  • Introduced Tab completions, Composer multi-file editing, and agent mode as core editor features, not plugins
  • Credit-based pricing since June 2025: Hobby (free), Pro ($20/mo), Pro+ ($60/mo), Ultra ($200/mo), Teams ($40/user/mo)
  • Supports Claude Sonnet/Opus, GPT-4o/5.4, and Gemini — auto mode selects the optimal model per task
  • Over 1 million developers worldwide rely on Cursor as their primary code editor

The Founding Story of Cursor

From research thesis to the fastest-growing code editor in developer tools history.

Why Anysphere Built Cursor

In 2022, the founding team at Anysphere observed a fundamental disconnect in developer tooling. Large language models had reached a capability threshold where they could understand, generate, and debug code at a professional level — yet the tools developers used every day were designed for a pre-AI world. VS Code, the dominant editor with over 70% market share, offered AI only through third-party extensions constrained by a plugin API never designed for deep AI integration.

Anysphere's thesis was straightforward: fork VS Code's open-source core (Code OSS, MIT license) and rebuild the editing experience with AI as a first-class primitive. This meant integrating AI not as a sidebar or a suggestion popup, but as a fundamental part of how developers write, edit, navigate, and debug code. The result was Cursor — an editor that looks and feels like VS Code but thinks like a coding partner.

The founding team brought together expertise in machine learning, programming language theory, and developer experience. Their backgrounds spanned Stanford computer science research and industry experience at leading AI labs. Rather than building a general-purpose AI assistant, they focused exclusively on the code editing workflow — understanding that the nuances of multi-file editing, context management, and iterative debugging required purpose-built solutions.

Growth and Evolution

Cursor's early traction came from developers who were frustrated with the limitations of extension-based AI tools. GitHub Copilot, the market leader at the time, operated within VS Code's extension API — which meant it could suggest code inline but could not edit multiple files simultaneously, run terminal commands, or autonomously iterate on errors. Cursor's deeper integration solved all three problems.

The editor reached its first 100,000 users within months of public launch. Tab completions drove initial adoption: developers found that Cursor's multi-line predictions were materially better than single-line autocomplete because the model had access to the full editor context — open files, terminal output, project structure — rather than just the current buffer. Composer and agent mode arrived shortly after, enabling workflows that simply were not possible in any other editor.

By 2025, Cursor had crossed one million active developers. Anysphere introduced credit-based pricing in June 2025 to align costs with actual AI model usage. The Hobby plan remained free. Pro at $20/month included unlimited Tab completions and a monthly credit pool for premium models. Pro+ ($60/month) and Ultra ($200/month) served developers who used agent mode extensively. Teams ($40/user/month) added SSO, shared Cursor Rules, and centralized billing for engineering organizations.

Cursor Timeline — Key Milestones

A chronological overview of Cursor's development from founding to the present day.

YearMilestoneDetails
2022Anysphere FoundedSan Francisco-based AI company established with the mission to build the AI-first IDE
2022VS Code ForkCursor forked from Code OSS (MIT license), preserving full extension and theme compatibility
2023Public LaunchCursor released for macOS, Windows, and Linux with Tab completions and inline AI chat
2023Composer IntroducedMulti-file editing from natural language instructions with diff preview and undo support
2024Agent Mode LaunchAutonomous coding — write files, run commands, install packages, iterate on errors
20241 Million DevelopersCursor crossed one million active users, becoming the fastest-growing code editor
2024@Codebase FeatureProject-wide indexing for context-aware answers about any file in the repository
2025Credit-Based PricingMonthly credit pools replaced fixed request limits — Auto mode unlimited, premium models from pool
2025MCP Server SupportModel Context Protocol integration for connecting AI to external tools and data sources
2025Cursor RulesProject-level .cursorrules files for team-wide AI behavior configuration
2026Enterprise TierCustom pricing with dedicated infrastructure, advanced SSO, audit logs, and priority support

Cursor's Core Philosophy

Three principles guide every product decision at Anysphere.

AI as a First-Class Primitive

AI is not a feature in Cursor — it is the foundation. Every editing action, from Tab completions to multi-file refactors, is designed around the assumption that an AI model is always available. This is why Cursor forked VS Code rather than building an extension: deep integration requires control over the entire editor stack, from the text buffer to the terminal to the file system.

Developer Control at Every Step

Cursor never applies changes without developer review. Composer shows diffs before writing. Agent mode displays every step in the chat panel and pauses for approval on destructive operations. Tab completions appear inline — accept with Tab, reject with Escape. The AI accelerates the workflow; the developer retains full authority over the codebase.

VS Code Compatibility

Switching editors is a high-friction decision. Cursor eliminates that friction by maintaining full compatibility with VS Code extensions, themes, keybindings, and workspace settings. Import your VS Code profile in one click. Every extension from the marketplace works natively. Developers get AI capabilities without abandoning the tools and configurations they already depend on.

The Technology Behind Cursor

How Cursor's architecture enables deeper AI integration than extension-based approaches.

Multi-Model Architecture

Cursor does not rely on a single AI provider. The editor integrates with Claude Sonnet and Opus from Anthropic, GPT-4o and GPT-5.4 from OpenAI, and Gemini from Google. Auto mode — available on all paid plans at zero credit cost — analyzes each request and routes it to the model best suited for the task. Simple completions go to fast models. Complex multi-file refactors route to frontier models with larger context windows.

This multi-model approach means Cursor is not dependent on any single provider's availability or pricing changes. When a new model launches — whether from Anthropic, OpenAI, Google, or another provider — Anysphere evaluates it and adds it to the model roster within days. Developers benefit from competition between model providers without managing API keys or configurations.

Local Indexing and Context

The @codebase feature relies on a local index of the developer's project. When a project is opened in Cursor, the editor scans every file and builds a semantic index stored on the developer's machine. This index powers context-aware completions — the AI understands not just the current file but the entire project structure, including imported modules, type definitions, test files, and configuration.

Local indexing is a deliberate architectural choice for both performance and privacy. Queries against the index execute in milliseconds with no network round-trip. The index never leaves the developer's machine unless explicitly shared in a chat query. This approach satisfies enterprise security requirements while delivering the context depth that makes Cursor's AI suggestions meaningfully better than tools limited to the current file buffer.

Frequently Asked Questions About Cursor

Common questions about Cursor's history, technology, and how it compares to other code editors.

Who founded Cursor?

Cursor was founded in 2022 by Anysphere, a San Francisco-based AI company. The founding team includes researchers and engineers with backgrounds in machine learning, programming language theory, and developer experience. Their mission was to build the IDE of the future by making AI a first-class citizen in the code editing workflow.

Is Cursor a fork of VS Code?

Yes. Cursor is built on VS Code's open-source codebase (Code OSS, MIT license). Every VS Code extension, theme, keybinding, and workspace feature works natively. The fork allows Anysphere to integrate AI at the editor core — Tab completions, Composer, agent mode — which is not possible through the standard extension API.

How does Cursor differ from GitHub Copilot?

GitHub Copilot is a VS Code extension providing inline suggestions. Cursor is an entire IDE with AI at its core. Cursor offers Composer for multi-file editing, agent mode for autonomous coding, @codebase for project-wide context, and Cursor Rules for project-specific AI behavior. The deeper integration enables workflows not possible with extension-based tools.

What AI models does Cursor use?

Cursor integrates Claude Sonnet and Opus (Anthropic), GPT-4o and GPT-5.4 (OpenAI), and Gemini (Google). Auto mode selects the best model per task at zero credit cost. Paid plans allow manual model selection from the monthly credit pool. The model roster updates as new versions become available.

Is Cursor open source?

Cursor itself is proprietary, but it is built on VS Code's open-source Code OSS (MIT license). The AI features — Tab completions, Composer, agent mode — are proprietary to Anysphere. All open-source VS Code extensions and themes work natively in Cursor.