Background Interventions found to be effective in research settings are often not as effective when implemented in community settings. Considering children with autism, studies have rarely examined the efficacy of laboratory‐tested interventions on child outcomes in community settings using randomized controlled designs. Methods One hundred and thirteen children with autism enrolled in public early intervention classrooms in low resource neighborhoods were randomized to Joint Attention, Symbolic Play, Engagement, and Regulation (JASPER) intervention or treatment as usual waitlist for 10 weeks with 1‐month follow‐up. Results Teaching assistants delivered JASPER at adequate fidelity. Children randomized to JASPER demonstrated significant gains over treatment as usual in core developmental outcomes of joint engagement, joint attention, and play skills that were maintained at follow‐up. Conclusions Supervised teaching assistants delivered JASPER intervention with a range of toddlers with autism leading to significant gains in developmental outcomes.