Introduction Social skills training interventions for children with autism spectrum disorder (ASD) typically focus on a skills deficit model rather than building on existing skills or encouraging the child to seek their own solutions. LEGO-based therapy is a child-oriented intervention to help improve social interactional skills and reduce isolation. The therapy is designed for school-age children with ASD and uses group-based play in a school setting to encourage peer relationships and social learning. Despite the reported potential benefits of LEGO-based therapy in a prior randomised controlled trial (RCT) and its adoption by many schools, the evidence to support its effectiveness on the social and emotional well-being of children with ASD is limited and includes no assessment of cost-effectiveness. Methods and analysis This multicentre, pragmatic, cluster RCT will randomise 240 participants (aged 7–15 years) with a clinical diagnosis of ASD to receive usual care or LEGO-based therapy with usual care. Cluster randomisation will be conducted on a school level, randomising each school as opposed to each individual child within a school. All prospective participants will be screened for eligibility before assenting to the study (with parents giving informed consent on behalf of their child). All participants will be followed up at 20 and 52 weeks after randomisation to assess for social, emotional and behavioural changes. The primary outcome measure is the social skills subscale of the Social Skills Improvement System completed by a teacher or teaching assistant associated with participating children at the 20-week follow-up time point. Ethics and dissemination Ethics approval has been obtained via the University of York Research Ethics Committee. The results of the trial will be submitted for publication in a peer-reviewed journal and will be disseminated to participating families, education practitioners and the third sector including voluntary and community organisations. Trial registration number ISRCTN64852382; Pre-results.