Getting Started with Cursor
What Is Cursor and the AI-Native IDE Philosophy
Cursor is a fork of VS Code built from the ground up around AI assistance. Learn the foundational philosophy behind AI-native IDEs and how Cursor differs from AI plugins in traditional editors.
The AI-Native IDE Concept
Traditional code editors like VS Code were designed as text editors first, with AI bolted on later via extensions like GitHub Copilot. Cursor takes the opposite approach: it was built with AI as a first-class architectural concern from day one.
Cursor is a fork of VS Code, so you get the familiar interface, extension ecosystem, and keybindings. But underneath, it has been restructured so that the AI has deep, native access to your codebase, not just the file you have open.
Why the Distinction Matters
When AI is a plugin, it operates as an afterthought — reading text in the editor buffer and injecting completions. It doesn't understand your project structure, your dependencies, or how files relate to each other.
When AI is native to the IDE, it can:
- Index your entire codebase and maintain a semantic understanding of it
- Understand cross-file relationships and call graphs
- Apply changes across multiple files in a single operation
- Know your project's conventions by analyzing the whole codebase
This architectural difference is what makes Cursor feel qualitatively different from Copilot, even when both use similar underlying models.
The Codebase as the Primary Context
The central theory in AI-native IDEs is that the codebase is the context. Traditional LLM interactions have a narrow context window — you paste code in, get code back. Cursor inverts this by continuously indexing your codebase and making it available as embedded context.
When you ask Cursor to implement a feature, it doesn't just look at the open file. It searches for relevant code across your project, brings that into context, and writes changes that are consistent with your existing patterns — automatically.
Cursor's Core Feature Set
| Feature | Description |
|---|---|
| Tab completion | Multi-line, context-aware completions that predict what you're about to write |
| Chat (Cmd+L) | Conversational AI with full codebase context |
| Composer (Cmd+I) | Multi-file editing agent for larger tasks |
| Agent mode | Autonomous task execution with terminal access |
| Codebase indexing | Semantic search across your entire project |
| .cursorrules | Project-level instructions that shape every AI interaction |
The Model Layer
Cursor is model-agnostic. You can configure which AI model powers each feature:
- Claude 3.5 Sonnet / Claude 3.7 Sonnet (Anthropic)
- GPT-4o / o1 (OpenAI)
- Gemini 1.5 Pro (Google)
- Cursor's own fine-tuned models for completions
This means Cursor is not a model — it's an AI-native editing environment that can be powered by the best available model for each task.
Key Takeaways
- Cursor is VS Code rebuilt with AI as a first-class architectural concern, not an add-on
- The codebase is the primary context — Cursor indexes your entire project, not just open files
- It is model-agnostic and can use Claude, GPT, Gemini, or Cursor's own models
- The philosophical shift from "AI plugin" to "AI-native IDE" changes what AI assistance can do