Background: Many countries are seeking to eliminate tuberculosis (TB), but incidence remains high in socially excluded groups such as people experiencing homelessness. There is limited research into the effectiveness of residential respite services (RRS), which provide accomodation and social and clinical support for homeless people with active TB. Methods: We used a register of all cases of TB diagnosed in London between 1 January 2010 and 3 October 2019 to compare characteristics and outcomes of patients treated in an RRS with patients receiving standard care. The primary outcome was successful treatment completion. We used logistic regression to compare likelihood of completing treatment, and simulation to estimate the absolute change in treatment completion resulting from this service. Results: A total of 78 homeless patients finished an episode of TB treatment at the RRS. Patients treated in the RRS were more likely than patients treated in standard care to have clinical and social risk factors including drug resistance, history of homelessness, drug or alcohol use, and need for directly observed therapy. After adjusting for these factors, patients treated in the RRS had 2.97 times the odds of completing treatment (95% CI = 1.44–6.96). Treatment ended in failure for 8/78 patients treated in the RRS (10%, 95% CI = 5%–20%). We estimated that in the absence of the RRS, treatment would have ended in failure for 17/78 patients (95% CI = 11–25). Conclusion: The residential respite service for homeless TB patients with complex social needs was associated with better treatment outcomes.