Introduction: Pre-exposure prophylaxis (PrEP) can provide high protection against HIV infection and is a recommended intervention for HIV-negative persons with substantial HIV risk, such as individuals with a partner living with HIV.  Demonstration projects of PrEP have been conducted in diverse settings worldwide to illustrate practical examples of how PrEP can be delivered.  Methods: We evaluated delivery of PrEP for HIV-negative partners within heterosexual HIV serodiscordant couples in an open-label demonstration project in East Africa.  The delivery model integrated PrEP into HIV treatment services, prioritizing PrEP for HIV-negative partners within serodiscordant couples prior to and during the first 6 months after the partner living with HIV initiated antiretroviral therapy (ART).  We measured adherence to PrEP through medication event monitoring system (MEMS) bottle caps and quantification of tenofovir in plasma among a random sample of participants. We estimated HIV infections prevented using a counterfactual cohort simulated from the placebo arm of a previous PrEP clinical trial. Results: We enrolled 1,010 HIV serodiscordant couples that were naïve to ART and PrEP.  Ninety-seven percent (97%) of HIV-negative partners initiated PrEP, and when PrEP was dispensed, objective measures suggest high adherence: 71% of HIV-negative participants took ≥80% of expected doses, as recorded via MEMS, and 81% of plasma samples had tenofovir detected.  A total of 4 incident HIV infections were observed (incidence rate=0.24 per 100 person-years), a 95% reduction (95% CI 86-98%, p<0.0001) in HIV incidence, relative to estimated HIV incidence for the population in the absence of PrEP integrated into HIV treatment services.   Conclusions: PrEP uptake and adherence were high and incident HIV was rare in this PrEP demonstration project for African HIV-negative individuals whose partners were known to be living with HIV.  Delivery of PrEP to HIV-negative partners within HIV serodiscordant couples was feasible and should be prioritized for wide-scale implementation.