Mental health problems are common among prison inmates. Music therapy has been shown to reduce mental health problems. It may also be beneficial in the rehabilitation of prisoners, but rigorous outcome research is lacking. We compared group music therapy with standard care for prisoners in a pilot randomised controlled trial that started with the establishment of music therapy services in a prison near Bergen in 2008. In all, 113 prisoners agreed to participate. Anxiety (STAI-State [State–Trait Anxiety Inventory], STAI-Trait), depression (HADS-D [Hospital Anxiety and Depression Scale]), and social relationships (Quality of Life Enjoyment and Satisfaction Questionnaire [Q-LES-Q]) were assessed at baseline; every 2 weeks in the experimental group; after 1, 3, and 6 months in the control group; and at release. No restrictions were placed on the frequency, duration, or contents of music therapy. Duration of stay in the institution was short (62% stayed less than 1 month). Only a minority reached clinical cutoffs for anxiety and depression at baseline. Between-group analyses of effects were not possible. Music therapy was well accepted and attractive among the prisoners. Post hoc analysis of within-group changes suggested a reduction of state anxiety after 2 weeks of music therapy (d = 0.33, p = .025). Short sentences and low baseline levels of psychological disturbance impeded the examination of effects in this study. Recommendations for planning future studies are given, concerning the careful choice of participants, interventions and settings, comparison condition and design aspects, choice of outcomes, and integration of research approaches. Thus, the present study has important implications for future studies evaluating interventions for improving prisoners’ mental health. Trial registration: ISRCTN22518605