Developers can now run agents on user-managed machines to grant them access to internal services or specialized hardware while keeping the agent loop in Cursor.
Teams with strict security constraints or specialized compute needs can now deploy Cursor agents inside their own private networks.
![]()