Cursor Launches Cloud AI Agents That Autonomously Build and Test Code, Already Driving 30% of Merged Pull Requests

Cursor launches cloud AI agents that autonomously build, test, and validate code in isolated virtual machines, already responsible for over 30% of merged pull requests at the company — signaling a major shift toward fully self-driving software development.

• Cursor has launched cloud agents that run in isolated virtual machines, enabling them to autonomously build, test, and validate software changes before producing merge-ready pull requests with video, screenshot, and log artifacts.
• Cloud agents are accessible across web, mobile, desktop, Slack, and GitHub, and can handle complex tasks like building new features, reproducing security vulnerabilities, fixing UI bugs, and conducting full site walkthroughs without competing for local resources.
• Over 30% of merged PRs at Cursor are now created autonomously by cloud agents, signaling a major shift toward self-driving codebases where agents will eventually manage merges, rollouts, and production monitoring end-to-end.