We implemented community-based direct observation of treatment, short course (DOTS), including a randomized controlled trial of direct observation either by community health workers (CHWs) or family members, under operational conditions in a region of Swaziland. There was a high death rate of 15%, due to the high HIV rates in the region. There was no significant difference in the cure and completion rate between direct observation of treatment by CHWs and family members [2% difference (95% CI -3% to 7%), exact P = 0.52]. A before-and-after comparison of outcomes demonstrated that the cure and treatment completion rate improved from a baseline of 27-67% following implementation of community-based DOTS. We conclude that community-based tuberculosis DOTS can improve successful outcomes of treatment. However, direct observation can be undertaken effectively using either daily family or CHW supervision. The choice of treatment supporter should be based on access, patient preference and availability of CHW resource.