It is estimated that one-third of the world's population is infected with Mycobacterium tuberculosis. Infection typically remains latent, but it can reactivate to cause clinical disease. The only vaccine, Mycobacterium bovis bacillus Calmette-Guérin (BCG), is largely ineffective, and ways to enhance its efficacy are being developed. Of note, the candidate booster vaccines currently under clinical development have been designed to improve BCG efficacy but not prevent reactivation of latent infection. Here, we demonstrate that administering a multistage vaccine that we term H56 in the adjuvant IC31 as a boost to vaccination with BCG delays and reduces clinical disease in cynomolgus macaques challenged with M. tuberculosis and prevents reactivation of latent infection. H56 contains Ag85B and ESAT-6, which are two of the M. tuberculosis antigens secreted in the acute phase of infection, and the nutrient stress-induced antigen Rv2660c. Boosting with H56/IC31 resulted in efficient containment of M. tuberculosis infection and reduced rates of clinical disease, as measured by clinical parameters, inflammatory markers, and improved survival of the animals compared with BCG alone. Boosted animals showed reduced pulmonary pathology and extrapulmonary dissemination, and protection correlated with a strong recall response against ESAT-6 and Rv2660c. Importantly, BCG/H56-vaccinated monkeys did not reactivate latent infection after treatment with anti-TNF antibody. Our results indicate that H56/IC31 boosting is able to control late-stage infection with M. tuberculosis and contain latent tuberculosis, providing a rationale for the clinical development of H56.