Cursor's long-running agents research preview is now available to all Ultra, Teams, and Enterprise users. These autonomous agents can work on ambitious projects for hours or days, producing large PRs with high merge rates. The system uses custom harnesses that enable planning before execution and multiple agents checking each

5m read time From cursor.com
Post cover image
Table of contents
# Making models more capable# Findings to date# Using long-running agents at Cursor# Toward self-driving codebases

Sort: