We aimed to investigate whether the Good School Toolkit reduced emotional violence, severe physical violence, sexual violence and injuries from school staff to students, as well as emotional, physical and sexual violence between peers, in Ugandan primary schools. We performed a two-arm cluster randomised controlled trial with parallel assignment. Forty-two schools in one district were allocated to intervention (n = 21) or wait-list control (n = 21) arms in 2012. We did cross-sectional baseline and endline surveys in 2012 and 2014, and the Good School Toolkit intervention was implemented for 18 months between surveys. Analyses were by intention to treat and are adjusted for clustering within schools and for baseline school-level proportions of outcomes. The Toolkit was associated with an overall reduction in any form of violence from staff and/or peers in the past week towards both male (aOR = 0.34, 95%CI 0.22-0.53) and female students (aOR = 0.55, 95%CI 0.36-0.84). Injuries as a result of violence from school staff were also lower in male (aOR = 0.36, 95%CI 0.20-0.65) and female students (aOR = 0.51, 95%CI 0.29-0.90). Although the Toolkit seems to be effective at reducing violence in both sexes, there is some suggestion that the Toolkit may have stronger effects in boys than girls. The Toolkit is a promising intervention to reduce a wide range of different forms of violence from school staff and between peers in schools, and should be urgently considered for scale-up. Further research is needed to investigate how the intervention could engage more successfully with girls.