Cursor launches Projects, letting one coordinator manage thousands of subagents

On September 10, AI coding tool Cursor officially launched Projects, a new feature aimed at helping developers take on larger bodies of work, such as building a feature, running a migration, or developing an entire application. With Projects, developers talk to a coordinator agent that delegates tasks to thousands of subagents. The coordinator does not write code itself, so it is never blocked and remains responsive to direction.

The feature combines cloud-by-default execution with local testing. A Project runs on its own computer, so closing a laptop does not stop it, and it can run more subagents in parallel than a local machine could support. Projects also maintains shared context across cloud and local machines, letting agents accumulate research, codebase knowledge, and user preferences over time. A subscription system lets the coordinator watch Slack channels, follow schedules, or track pull requests, taking action automatically when CI fails or PRs are opened or merged.

Cursor says it has been using Projects internally for months, completing migrations of hundreds of PRs, maintaining design-system consistency, and shipping Projects itself. According to the company, new users merge 30% more PRs, while users who primarily use Projects merge six times as many. Projects is now available in beta and is rolling out to all users starting September 10, 2026.