Many campuses have utilized linked-course communities in an effort to enhance learning and build community, but most of the research on these communities are case studies or correlational designs subject to selection effects. This study conducted a randomized controlled trial of STEM linked-course communities for first-semester students with STEM majors. Students (N = 291) were randomly assigned to either a linked-course community (wherein students took three courses with the same peers) or a control group (wherein students took similar courses but with a random mix of students). Utilizing institutional records, student progress was tracked for the students' first year. Results showed that students in the community were retained into the sophomore year 10 percentage points higher at the university and 15 percentage points higher in STEM, compared to the control group. Additional comparisons to the control group showed that community students earned more credits in both the fall and spring, more STEM credits in the spring, and higher grades in the fall. The effects were consistent across a variety of demographic groups (race, first-generation status, gender, etc.), though the retention effects were significantly stronger for students with low math placement scores. Mediation analyses showed that, compared to the control group, students in the communities had a higher number of classes with a peer, which in turn predicted increased retention. Given that this seems to be the first randomized controlled trial of linked-course communities utilizing students' majors, future research should replicate these findings and more deeply explore the mechanisms by which these communities increase student success.