Cursor is an AI coding agent for building software. You describe a task, and agents write, test and change code in your repositories, so you can focus on the decisions that matter.
The page describes agents that run in parallel and can work for hours or days on larger tasks. Cloud agents use their own computers to build and test features end to end, then hand back work for review. Cursor also runs in the terminal through its CLI, works in Slack, and reviews pull requests on GitHub.
The product is aimed at software teams. The page names teams that build products and says Cursor is used across large organisations, and it lists enterprise features such as single sign-on, audit logs, spend controls and access controls for repositories and models.
Cursor lets you choose from several AI models, including models from OpenAI, Anthropic and Google, and it offers an Auto mode that picks one for each task. Plans are sold per person, with a free plan and paid plans billed monthly. Third-party model usage is billed separately by usage, and the page says Enterprise customers can pay by invoice.
Cursor is available as a download for macOS, and the team behind it is Anysphere, Inc. The page also describes a research team working on the future of software development and publishes a changelog and technical reports.
