Understanding mechanisms and active ingredients of intervention is critical to training clinicians, particularly when interventions are transported from laboratories to communities. One promising active ingredient of parenting programs is clinicians' in vivo feedback regarding parent-child interactions. The present study examined whether a form of in vivo feedback, in the moment commenting, predicted treatment retention and parent behavior change when the Attachment and Biobehavioral Catch-up (ABC) intervention was implemented in a community setting. Observational data were collected from 78 parent-child dyads (96% mothers; M age = 29 years; 81% minority; infants' M age = 12 months; 90% minority) across 640 sessions conducted by 9 clinicians (100% female, M age = 39; 67% minority) in Hawaii. Parental behavior was assessed with a semistructured play task before and after intervention. Clinicians' in-the-moment feedback to parents was assessed from intervention session videos. Clinicians' frequency and quality of in-the-moment feedback predicted change in parental intrusiveness and sensitivity at posttreatment. Frequency of in-the-moment feedback also predicted likelihood of retention. Hierarchical linear modeling demonstrated strong support for these associations at the between-clinician level, and limited additional support at the within-clinician (i.e., between-case) level. Thus, a hypothesized active ingredient of treatment, in-the-moment feedback, predicted community-based ABC outcomes. The results complement lab-based evidence to suggest that in vivo feedback may be a mechanism of change in parenting interventions. Helping clinicians to provide frequent, high-quality in vivo feedback may improve parenting program outcomes in community settings.