OBJECTIVE: To describe the approach used to promote adherence to antiretroviral therapy (ART) and to present the outcomes in the first primary care public sector ART project in South Africa. DESIGN: The study is a prospective open cohort, including all adult patients naive to previous ART who received antiretroviral treatment in Khayelitsha, from May 2001 to the end of 2002. Patients were followed until their most recent visit before 31 July 2003. METHODS: Plasma viral load was determined at 3, 6, 12, 18 and 24 months after ART was initiated, and CD4 cell counts 6-monthly. Kaplan-Meier estimates were determined for the cumulative proportions of patients surviving, and patients with viral load suppression and viral rebound. RESULTS: A total of 287 patients were initiated on triple therapy. The probability of survival was 86.3% at 24 months. The median CD4 cell count gain was 288 cells/microliters at 24 months. Viral load was less than 400 copies/ml in 89.2, 84.2 and 69.7% of patients at 6, 12 and 24 months, respectively. The cumulative probability of viral rebound (two consecutive HIV-RNA measurements above 400 copies/ml) after achieving an HIV-RNA measurement below 400 copies/ml was 13.2% at 18 months. CONCLUSION: The study shows that, with a standard approach to patient preparation and strategies to enhance adherence, a cohort of patients on ART can be retained in a resource-limited setting in a developing country. A high proportion of patients achieved suppression of viral replication. The subsequent probability of viral rebound was low.