Cursor: The AI Code Editor That Forced GitHub Copilot to Respond

Most developer tools grow slowly through enterprise sales cycles, conference demos, and gradual word-of-mouth adoption. Cursor grew differently. In 2024, it became the most-discussed new developer tool in the industry — not because of a marketing campaign or a major partnership, but because experienced developers who tried it publicly said it changed how they worked in ways that GitHub Copilot had not. Understanding what Cursor does differently, and where it falls short, explains why it became a significant competitor to the most established AI coding tool in the market. ## What Cursor Is Cursor is a code editor — specifically, a fork of Visual Studio Code — built by Anysphere Inc., founded in 2022. It is not a plugin for an existing editor. It is a full editor, built from the VS Code source code, with AI capabilities integrated at every level of the interface rather than added on top of it. This architectural choice is the source of both Cursor's main strength and its most common objection. Anysphere raised a $60 million Series A in August 2023, led by Andreessen Horowitz with participation from OpenAI. The OpenAI involvement is notable — it means one of GitHub Copilot's core technology partners also invested in its main competitor. By 2024, multiple sources reported Cursor had crossed $100 million in annual recurring revenue, a pace of growth that few developer tools have matched. ## The Core Difference: Codebase-Aware AI GitHub Copilot, in its standard configuration, sees the file you are working in plus a limited amount of surrounding context from recently opened files. This is effective for generating individual functions and boilerplate, but it means Copilot cannot understand how the code you are writing fits into the larger system it will become part of. Cursor's defining feature is codebase indexing. When you open a project in Cursor, it indexes the entire repository — every file, every function, every variable name and type signature — and stores that index locally. When you ask Cursor's AI anything, it can retrieve relevant pieces of the codebase from that index to inform its response. In practice, this means you can ask Cursor "how does the authentication flow work in this project?" and receive an answer that cites specific files and functions. You can ask "write a new endpoint that follows the same pattern as the existing user endpoints" and Cursor will look at those existing endpoints before generating code. You can ask "why is this test failing?" and Cursor can trace the failure through imports and dependencies across multiple files. This is qualitatively different from what a context-limited inline completion tool can do, and it is the primary reason experienced developers who work on large codebases report that Cursor changed their workflow in a way that Copilot did not. ## Key Features **Tab completion.** Cursor's inline completion (called Tab) is similar to GitHub Copilot's inline suggestions. It predicts what you are about to type and completes it. Cursor has added a capability called "next edit prediction" — after accepting a completion, it predicts where your cursor will move next and what you will edit there, reducing the number of keystrokes needed to make a set of related changes. **Cursor Chat.** The chat panel can reference the codebase (@codebase), specific files (@filename), web URLs (@url), and documentation for popular libraries (@docs). This makes Cursor's chat meaningfully more capable than a general-purpose AI chat for coding questions, because it has access to the specific context of your project. **Composer.** Cursor's most powerful feature is Composer (rebranded to Agent in newer versions), which allows you to describe a multi-file change in natural language and have Cursor plan and execute it. Composer can create new files, modify multiple existing files, run terminal commands, and iterate based on the output — all within a single session. This "agentic" mode is where Cursor most clearly surpasses what Copilot offers as of 2024. **Model selection.** Cursor lets users choose which AI model powers their experience. Claude 3.5 Sonnet (Anthropic), GPT-4 Turbo (OpenAI), and Cursor's own cursor-small model (optimised for speed on simpler completions) are all available. This flexibility matters because different models excel at different tasks — Claude 3.5 Sonnet has been particularly strong at multi-file reasoning and instruction following, which aligns with Cursor's strengths. ## Strengths **Superior codebase understanding.** For experienced developers working on complex, large codebases, the indexed codebase access is the most significant practical differentiator among coding AI tools as of 2024. Asking questions about how the system works and getting answers grounded in the actual code, rather than in generic programming knowledge, changes the utility of the AI. **Multi-file editing.** Composer's ability to plan and execute changes across multiple files simultaneously handles a category of task — refactoring, adding a feature that requires changes in several places, updating a pattern consistently throughout a codebase — that inline completion tools are not designed for. **Model flexibility.** Using the best available model for a given task, rather than being tied to a single provider's model, gives Cursor an adaptability advantage as the AI model landscape continues to evolve. ## Weaknesses **Requires switching editors.** There is no Cursor plugin for VS Code. Cursor is a separate editor. For developers who have spent years customising their VS Code environment, have extensions that do not have equivalents in Cursor, or work in organisations that standardise on a specific tool, switching to Cursor requires a meaningful commitment. JetBrains users are not served at all. This is the single most common objection to Cursor adoption. **Privacy and code confidentiality.** Cursor's functionality — particularly codebase indexing and Composer — requires sending code to Cursor's servers for processing. Cursor offers a Privacy Mode setting that prevents code from being stored or used for training, and a Business tier with stronger data handling guarantees, but organisations in regulated industries (finance, healthcare, defence) that require air-gapped or on-premises AI tools cannot use Cursor in its current form. **Maturity and stability.** Cursor is a young product. Features change between releases, keybindings and interface behaviours are less settled than in a mature editor, and the product has shipped bugs that affect daily workflows. Developers who need a stable, predictable tooling environment may find Cursor's pace of change disruptive. **Enterprise ecosystem.** GitHub Copilot has deep integration with GitHub's enterprise infrastructure — SSO, audit logs, policy management, usage analytics. Cursor's enterprise offering is less developed. Large organisations evaluating AI coding tools often need these procurement and compliance capabilities before they can deploy at scale. ## Pricing - **Free**: 2,000 completions, 50 premium model requests, limited Composer usage per month. - **Pro**: $20/month — 500 premium model requests (GPT-4, Claude), unlimited completions, more Composer usage. - **Business**: $40/user/month — enforced privacy mode, centralised billing, usage dashboard. ## Who Cursor Is Best For Cursor is best for individual developers and small teams who work on complex codebases, are comfortable with VS Code as their base, want the most capable multi-file AI editing currently available, and are willing to accept the privacy tradeoffs of a cloud-based tool. It is the right choice for developers for whom codebase understanding matters more than ecosystem integration — and for whom switching editors is a worthwhile investment.