Introduction: Tuberculosis is difficult to diagnose and it is the leading cause of death in HIV infected individuals in developing countries. There is an urgent need of low-cost diagnostic markers for resource-limited settings. Methods: The study involved 1571 patients from an HIV cohort study in India with known serum albumin concentrations at the time of becoming eligible for antiretroviral therapy (ART). We investigated the diagnostic accuracy of serum albumin to predict tuberculosis within six months of ART eligibility and the prognostic value in patients who experienced tuberculosis. Results: The diagnostic accuracy of serum albumin, measured by the area under the receiver operating characteristic curve, to predict tuberculosis was 0.81 (95% confidence interval 0.78-0.83). Serum albumin concentrations <3.2 g/dL were associated with 85% specificity and <4.1 g/dL were associated with negative predictive values >90%, even in settings with high tuberculosis prevalence. Hypoalbuminemia was associated with an increased risk of mortality in patients with tuberculosis. Conclusion: Serum albumin can be a useful low-cost diagnostic marker for tuberculosis in HIV infected patients eligible for ART. However, we failed to find thresholds to rule out or rule in tuberculosis. If these results are confirmed by other studies, serum albumin could be used to improve the diagnostic accuracy of intensive case finding algorithms for HIV-related tuberculosis. In patients who experience tuberculosis, hypoalbuminemia is associated with poor prognosis.