AI coding startup Cursor has announced an update to its AI coding agent, CNBC reported on Feb. 24.
The company said the update lets the agent test changes on its own and record its work with video, logs and screenshots. Users can run the agent on the web, the Cursor desktop app, mobile devices, the messaging platform Slack and Microsoft GitHub.
The agent runs in parallel in cloud-based virtual machines with independent development environments. The company said there is no need to share a developer laptop’s resources or waste time on initial setup.
The company said that while it previously handled 1 to 3 tasks at the same time, it can now run 10 to 20 simultaneously. It said throughput will rise sharply.
The Cursor agent can also take over complex tasks and repeatedly test and revise its work until a feature is complete. The company said this can reduce the time developers spend editing files and analyzing code, allowing them to focus on judgement and setting direction.