Several jurisdictions in the United States have secured hotels to temporarily house people experiencing homelessness who require isolation or quarantine for confirmed or suspected coronavirus disease 2019 (COVID-19). To our knowledge, little is known about how these programs serve this vulnerable population outside the hospital setting. To assess the safety of a hotel-based isolation and quarantine (I/Q) care system and its association with inpatient hospital capacity. This retrospective cohort study of a hotel-based I/Q care system for homeless and unstably housed individuals in San Francisco, California, was conducted from March 19 to May 31, 2020. Individuals unable to safely isolate or quarantine at home with mild to moderate COVID-19, persons under investigation, or close contacts were referred from hospitals, outpatient settings, and public health surveillance to 5 I/Q hotels. Of 1009 I/Q hotel guests, 346 were transferred from a large county public hospital serving patients experiencing homelessness. A physician-supervised team of nurses and health workers provided around-the-clock support, including symptom monitoring, wellness checks, meals, harm-reduction services, and medications for opioid use disorder. Characteristics of I/Q hotel guests, program retention, county hospital readmissions, and mean length of stay. Overall, the 1009 I/Q hotel guests had a median age of 44 years (interquartile range, 33-55 years), 756 (75%) were men, 454 (45%) were Latinx, and 501 (50%) were persons experiencing sheltered (n = 295) or unsheltered (n = 206) homelessness. Overall, 463 (46%) received a diagnosis of COVID-19; 303 of 907 (33%) had comorbid medical disorders, 225 of 907 (25%) had comorbid mental health disorders, and 236 of 907 (26%) had comorbid substance use disorders. A total of 776 of 955 guests (81%) completed their I/Q hotel stay; factors most strongly associated with premature discontinuation were unsheltered homelessness (adjusted odds ratio, 4.5; 95% CI, 2.3-8.6; P < .001) and quarantine status (adjusted odds ratio, 2.6; 95% CI, 1.5-4.6; P = .001). In total, 346 of 549 patients (63%) were transferred from the county hospital; of 113 ineligible referrals, 48 patients (42%) had behavioral health needs exceeding I/Q hotel capabilities. Thirteen of the 346 patients transferred from the county hospital (4%) were readmitted for worsening COVID-19. Overall, direct transfers to I/Q hotels from emergency and outpatient departments were associated with averting many hospital admissions. There was a nonsignificant decrease in the mean hospital length of stay for inpatients with confirmed or suspected COVID-19 from 5.5 to 2.7 days from March to May 2020 (P = .11). To support persons experiencing homelessness during the COVID-19 pandemic, San Francisco rapidly and safely scaled a hotel-based model of I/Q that was associated with reduced strain on inpatient capacity. Strategies to improve guest retention and address behavioral health needs not met in hotel settings are intervention priorities.