What Is Cursor? The AI Code Editor Explained

You open a project, type "add input validation to the signup form and write a test for it," and a few moments later the editor has read the relevant files, made the edits across several locations, run the test suite in the terminal, and shown you a diff to approve. No tab-by-tab navigation, no copy-pasting from a chat window. That is the pitch behind Cursor: an editor where an AI agent does the legwork and you stay in the review seat.

This breakdown covers what Cursor actually is, the company building it, the core features that set it apart, which models you can pick, what the four pricing tiers include, how Privacy Mode works, and the honest limitations worth knowing before you commit. Pricing and model figures below are reported by Cursor and were checked on June 9, 2026. Always confirm the current numbers on Cursor's pricing page before you pay.

What Is Cursor

Cursor is an AI code editor, sometimes described as an agent-native IDE, built by the company Anysphere. In practical terms it is a code editor that puts an AI agent at the center of the workflow: you write instructions in natural language, and the agent searches your codebase, edits files, runs terminal commands, and works through multi-step tasks on its own, pausing for you to review the changes.

The distinction worth drawing is between autocomplete and agency. Plenty of tools suggest the next line of code as you type. Cursor does that too, but its larger ambition is the agent: a system you can hand a whole task to, not just a line. You describe what you want, and instead of returning a snippet to paste, Cursor goes and makes the changes across the project, then hands you a diff.

It runs as a desktop editor and also reaches into the terminal through a command-line interface, with agents that can run on the web, on mobile, and in the cloud. Because Cursor supports the Model Context Protocol, it can connect to external tools and data sources through a shared open standard rather than one-off integrations. For where Cursor sits among other developer AI tools, see the AI Tools Hub.

The Company Behind Cursor

Cursor is built by Anysphere, Inc., which also trades under the Cursor name and is based in San Francisco. The company was founded in 2022 by Michael Truell, Sualeh Asif, Arvid Lunnemark, and Aman Sanger, who met as students at MIT. By 2025 the company had grown to roughly 300 employees. One of the founders, Arvid Lunnemark, left in October 2025 to start a safety lab called Integrous Research.

The funding story is steep, and the figures below come from independent reporting rather than the company itself. Cursor raised an $8M seed from the OpenAI Startup Fund in October 2023, then a $60M Series A at a $400M valuation, a $900M Series C in June 2025 that valued it at $9.9B, and a $2.3B Series D in November 2025 co-led by Accel and Coatue, with Google and Nvidia participating, that put its valuation at $29.3B. Revenue is reported to have climbed in step, reaching a reported $3B in annual recurring revenue by early 2026.

One development sits in a category of its own. In April 2026, xAI was reported to have secured the right to acquire Cursor for around $60B later in 2026, or to pay $10B for joint work instead. Treat this as reported and in progress rather than a closed deal. We flag it because, if it happens, it would reshape the ownership and direction of the product covered here.

Core Features at a Glance

The agent is the headline, but Cursor is several connected pieces. Here is what each one does and why it matters in day-to-day work.

Agent and parallel agents

The Agent is the core of Cursor. It searches the codebase, edits files, runs terminal commands, and completes multi-step tasks from a natural-language request. With Cursor 2.0 (October 2025), you can run multiple agents in parallel, each isolated in its own git worktree or on a remote machine, so two tasks can progress at once without stepping on each other's files.

Cursor Tab and Fusion

Cursor Tab is the fast autocomplete layer. Beyond finishing the current line, it predicts your next action, including "cursor jumps" that anticipate where you will move next in the file. It is powered by Cursor's in-house Fusion model, announced in January 2025.

Composer

Composer is Cursor's own low-latency agentic coding model, trained with codebase-wide semantic search and built to be deeply integrated with the editor. It has iterated quickly: Composer arrived in October 2025, followed by Composer 1.5, then Composer 2 (built on a Kimi K2.5 base), and the current Composer 2.5 as of May 2026.

Bugbot, MCP, and the CLI

Bugbot is an agentic code-review tool that plugs into GitHub pull requests, inspecting diffs and flagging issues; it launched in July 2025. Cursor supports the Model Context Protocol for connecting external tools, and the Cursor CLI brings the agent into the terminal with a Shell Mode and headless or CI usage, alongside integrations with GitHub, GitLab, JetBrains, Slack, Linear, and Xcode.

Codebase indexing and rules

Cursor builds an index of your project for codebase-wide semantic search, which is what lets the agent find the right files without you naming them. You can also customize its behavior with rules, skills, and prompts to match a team's conventions, and share those through a team marketplace. The exact configuration file names are part of Cursor's documentation rather than something to memorize here; the capability is what matters.

Which Models You Can Use

Cursor is model-flexible by design. You can choose between leading models from OpenAI, Anthropic, Gemini, and xAI, or Cursor's own, mixing third-party frontier models with the in-house ones. The model lineup moves fast, so treat the list below as a snapshot taken on June 9, 2026 and confirm the live roster in Cursor's documentation.

Two practical notes. First, many of these models support a Max Mode with up to a 1M-token context window, useful when you need the agent to reason across a large slice of a codebase at once. Second, the in-house models exist for a reason: Composer is tuned for low latency and tight editor integration, and Fusion is what makes Tab feel fast. For a deeper look at when to reach for which, see Cursor models explained.

Plans and Pricing at a Glance

Cursor sells four plans, all directly at cursor.com and with no resellers. The headline prices below are reported by Cursor and were verified on June 9, 2026. Pricing is usage-based underneath the subscription: every plan includes a set amount of model usage, and once you consume it, on-demand usage continues and is billed in arrears. The Individual tier shows a single $20 entry price and contains Pro, Pro+, and Ultra options; the higher option prices are not listed on the summary view, so confirm them on the pricing page.

A word on what "usage" means here, because it is where the surprises live. Your subscription buys a bucket of included model usage. Cross that line and you keep working on on-demand usage that is billed after the fact. Self-serve plans take major cards, while Enterprise bills by invoice or wire. If you want the tier-by-tier detail, including how the Individual options differ, see Cursor pricing explained and is Cursor free? for what the Hobby plan actually covers.

How Privacy Mode Works

For anyone working on proprietary code, the question is simple: does my code get used to train someone's model? Cursor's answer is Privacy Mode. It is a toggle in settings, and a team admin can enforce it across an entire organization so individual users cannot turn it off.

When Privacy Mode is on, Cursor guarantees that your code data is not used for training, either by Anysphere or by the third-party model providers it routes requests to. That is a meaningful, specific commitment. It is also worth being precise about its edges: the guarantee covers training. Storage and retention specifics are not spelled out in the summary sources, so if those details matter for your compliance posture, read Cursor's Security page directly rather than assuming.

Who Cursor Is For

Cursor fits a wide range of developers, but the value and the right plan shift by use case. Here is how the main groups line up.

Honest Limitations

Cursor is a strong tool, and the points below are not reasons to avoid it. They are reasons to use it with clear eyes, and two of them are the company's own missteps, which are more instructive than any feature list.

Frequently Asked Questions