AI-supported tools can aid programming learners by offering adaptive assistance, but research often centers on individual tools rather than broader design insights. A key design challenge is balancing learner autonomy with system guidance. To explore user needs, we conducted a participatory design study with 15 novice undergraduates and 10 instructors, followed by a survey of 172 introductory programming students. Learners favored help that is encouraging, visual, and peer-informed; instructors emphasized progress-based scaffolding and promoting best practices. Both groups preferred shared control, though learners leaned toward autonomy while instructors favored more guidance. Control preferences also varied by individual. We propose design guidelines for user-centered help features and adaptive control, contributing to more effective, human-centered AI-supported programming tools.