BACKGROUND: Tuberculosis control in sub-Saharan Africa has long been hampered by poor diagnostics and weak health systems. New molecular diagnostics, such as the Xpert® MTB/RIF assay, have the potential to improve patient outcomes. We present a cluster randomised trial designed to evaluate whether the positioning of this diagnostic system within the health system has an impact on important patient-level outcomes. METHODS/DESIGN: This pragmatic cluster randomised clinical trial compared two positioning strategies for the Xpert MTB/RIF system: centralised laboratory versus primary health care clinic. The cluster (unit of randomisation) is a 2-week time block at the trial clinic. Adult pulmonary tuberculosis suspects with confirmed human immunodeficiency virus infection and/or at high risk of multidrug-resistant tuberculosis are enrolled from the primary health care clinic. The primary outcome measure is the proportion of culture-confirmed pulmonary tuberculosis cases initiated on appropriate treatment within 30 days of initial clinic visit. Univariate logistic regression will be performed as the primary analysis using generalised estimating equations with a binomial distribution function and a logit link. CONCLUSION: Diagnostic research tends to focus only on performance of diagnostic tests rather than on patient-important outcomes. This trial has been designed to improve the quality of evidence around diagnostic strategies and to inform the scale-up of new tuberculosis diagnostics within public health systems in high-burden settings. TRIAL REGISTRATION: Current Controlled Trials ISRCTN18642314; South African National Clinical Trials Registry DOH-27-0711-3568.