Fewer than 20% of patients with multidrug-resistant (MDR) tuberculosis are receiving treatment and there is an urgent need to scale up treatment programmes. One of the biggest barriers to scale-up is the treatment regimen, which is lengthy, complex, ineffective, poorly tolerated and expensive. For the first time in over 50 years, new drugs have been developed specifically to treat tuberculosis, with bedaquiline and potentially delamanid expected to be available soon for treatment of MDR cases. However, if the new drugs are merely added to the current treatment regimen, the new regimen will be at least as lengthy, cumbersome and toxic as the existing one. There is an urgent need for strategy and evidence on how to maximize the potential of the new drugs to improve outcomes and shorten treatment. We devised eight key principles for designing future treatment regimens to ensure that, once they are proven safe in clinical trials, they will be clinically effective and programmatically practicable. Regimens should contain at least one new class of drug; be broadly applicable for use against MDR and extensively drug-resistant Mycobacterium tuberculosis complex strains; contain three to five effective drugs, each from a different drug class; be delivered orally; have a simple dosing schedule; have a good side-effect profile that allows limited monitoring; last a maximum of 6 months; and have minimal interaction with antiretrovirals. Following these principles will maximize the potential of new compounds and help to overcome the clinical and programmatic disadvantages and scale-up constraints that plague the current regimen.