Youth mentoring is a potentially powerful tool for prevention and intervention, but it has garnered little attention from clinical child and adolescent psychologists. For decades, the practice of youth mentoring has out-paced its underlying science, and meta-analytic studies consistently reveal modest outcomes. The field is now at an important crossroads: Continue to endorse traditional, widely used models of mentoring or shift to alternative models that are more in line with the tenets of prevention science. Presented here is a bilateral framework to guide the science and practice of mentoring going forward. Our premise is that mentoring relationships can serve as both means to a targeted end and as a valued end unto itself. We present a functional typology of current mentoring programs (supportive, problem-focused, & transitional) and call for greater specification of both the process and expected outcomes of mentoring. Finally, we argue that efforts to leverage mentoring relationships in service of youth development and the promotion of child and adolescent mental health will likely require disrupting the science, practice, and policy that surrounds youth mentoring.