Up to half the people with tuberculosis do not complete their treatment. Strategies to improve adherence to diagnostic and treatment regimens are therefore important. To assess the effects of directly observed therapy by an appointed agent (health worker, community volunteer, or family member) on cure and treatment completion in people on treatment for tuberculosis. We searched The Cochrane Controlled Trials Register, the Cochrane Infectious Diseases Group specialized trials register, MEDLINE, EMBASE, LILACS, and reference lists of articles. We also contacted experts in the field. Randomized and quasi-randomized trials of direct observation of tablet swallowing compared with self treatment for tuberculosis. Two reviewers independently assessed trial quality and extracted data. Four studies (n = 1603 participants) included. There was no difference detected between direct observation and self treatment for cure (Relative Risk [RR] 1.06, 95% confidence interval [CI] 0.98 to 1.14); and for cure plus treatment completion (RR 1.06, 95% CI 1.00 to 1.13). A stratified analysis by the appointed agent (health professional, lay health worker, or family/community member) did not reveal any important differences. One study conducted in an optimal setting in which patients were given a choice of supervisor did show modest benefit: cure (RR 1.13, 95% CI 1.04 to 1.24); cure plus treatment completion (RR 1.11, 95% CI 1.03 to 1.18). Randomized trials provide no evidence that directly observed therapy in low and middle income country settings improves cure or treatment completion rates in patients with tuberculosis.