Abstract Background The prevalence of depression in the elderly is growing worldwide, and the population aging in China makes depression a major health problem for the elderly adults and a tremendous burden to the society. Effective interventions should be determined to provide an approach solving the problem and improving the situation. This study examined the effectiveness of a mutual recovery program intervention on depressive symptom, sleep quality, and well-being in community-dwelling elderly adults with depressive symptom in Shanghai. Methods Recruitment was performed between July 2012 and August 2012. Using a cluster randomized wait-list controlled design, we randomized 6 communities (n = 237) into either the intervention group (3 communities, n = 105) or to a wait-list control group (3 communities, n = 132). All participants met the inclusion criteria for depression, which were defined by The Geriatric Depression Scale (GDS-15). From March to May of 2013, participants in the intervention group underwent a 2-month mutual recovery program intervention. The intervention included seven 90-min, weekly sessions that were based on a standardized self-designed schedule. Depression was used as primary outcome at three measurement moments: baseline (T1), before intervention at 24 weeks (T2), and immediately after intervention at 32 weeks (T3). Well-being and sleep quality were used as the secondary outcomes, and were evaluated based on the WHO-5 Well-being Index (WHO-5) and the Self-administered Sleep Questionnaire (SSQ). Finally, a total of 225 participants who completed all the sessions and the three measurements entered the final analysis. Mixed-model repeated measures ANOVAs were performed to estimate the intervention effects. Results There was no significant difference in gender, marriage, age structure, post-work type, and education background between the intervention and control group at baseline. Multivariate ANOVAs showed that there was no significant difference within the groups in terms of sleep, well-being, and depression at baseline and before the intervention. Mixed-model repeated measures ANOVAs detected a group × time interaction on depression, sleep, and well-being and showed a favorable intervention effect within groups immediately after the intervention. Conclusions The mutual recovery program could be a creative and effective approach to improve mental health in older community-dwelling adults with depressive symptom.