BACKGROUND: Four decades into the HIV epidemic, CNS infection remains a leading cause of preventable HIV-related deaths in routine care. The Driving Reduced AIDS-associated Meningo-encephalitis Mortality (DREAMM) project aimed to develop, implement, and evaluate pragmatic implementation interventions and strategies to reduce mortality from HIV-related CNS infection. METHODS: DREAMM took place in five public hospitals in Cameroon, Malawi, and Tanzania. The main intervention was a stepwise algorithm for HIV-related CNS infections including bedside rapid diagnostic testing and implementation of WHO cryptococcal meningitis guidelines. A health system strengthening approach for hospitals was adopted to deliver quality care through a co-designed education programme, optimised clinical and laboratory pathways, and communities of practice. DREAMM was led and driven by local leadership and divided into three phases: observation (including situational analyses of routine care), training, and implementation. Consecutive adults (aged ≥18 years) living with HIV presenting with a first episode of suspected CNS infection were eligible for recruitment. The primary endpoint was the comparison of 2-week all-cause mortality between observation and implementation phases. This study completed follow-up in September, 2021. The project was registered on ClinicalTrials.gov, NCT03226379. FINDINGS: From November, 2016 to April, 2019, 139 eligible participants were enrolled in the observation phase. From Jan 9, 2018, to March 25, 2021, 362 participants were enrolled into the implementation phase. 216 (76%) of 286 participants had advanced HIV disease (209 participants had missing CD4 cell count), and 340 (69%) of 494 participants had exposure to antiretroviral therapy (ART; one participant had missing ART data). In the implementation phase 269 (76%) of 356 participants had a probable CNS infection, 203 (76%) of whom received a confirmed microbiological or radiological diagnosis of CNS infection using existing diagnostic tests and medicines. 63 (49%) of 129 participants died at 2 weeks in the observation phase compared with 63 (24%) of 266 in the implementation phase; and all-cause mortality was lower in the implementation phase when adjusted for site, sex, age, ART exposure (adjusted risk difference -23%, 95% CI -33 to -13; p<0·001). At 10 weeks, 71 (55%) died in the observation phase compared with 103 (39%) in the implementation phase (-13%, -24 to -3; p=0·01). INTERPRETATION: DREAMM substantially reduced mortality from HIV-associated CNS infection in resource-limited settings in Africa. DREAMM scale-up is urgently required to reduce deaths in public hospitals and help meet Sustainable Development Goals. FUNDING: European and Developing Countries Clinical Trials Partnership, French Agency for Research on AIDS and Viral Hepatitis. TRANSLATIONS: For the French and Portuguese translations of the abstract see Supplementary Materials section.