We evaluated the safety and efficacy of fully closed-loop insulin therapy compared with standard insulin therapy in adults with type 2 diabetes requiring dialysis. In an open-label, multinational, two-center, randomized crossover trial, 26 adults with type 2 diabetes requiring dialysis (17 men, 9 women, average age 68 ± 11 years (mean ± s.d.), diabetes duration of 20 ± 10 years) underwent two 20-day periods of unrestricted living, comparing the Cambridge fully closed-loop system using faster insulin aspart ('closed-loop') with standard insulin therapy and a masked continuous glucose monitor ('control') in random order. The primary endpoint was time in target glucose range (5.6-10.0 mmol l-1). Thirteen participants received closed-loop first and thirteen received control therapy first. The proportion of time in target glucose range (5.6-10.0 mmol l-1; primary endpoint) was 52.8 ± 12.5% with closed-loop versus 37.7 ± 20.5% with control; mean difference, 15.1 percentage points (95% CI 8.0-22.2; P < 0.001). Mean glucose was lower with closed-loop than control (10.1 ± 1.3 versus 11.6 ± 2.8 mmol l-1; P = 0.003). Time in hypoglycemia (<3.9 mmol l-1) was reduced with closed-loop versus control (median (IQR) 0.1 (0.0-0.4%) versus 0.2 (0.0-0.9%); P = 0.040). No severe hypoglycemia events occurred during the control period, whereas one severe hypoglycemic event occurred during the closed-loop period, but not during closed-loop operation. Fully closed-loop improved glucose control and reduced hypoglycemia compared with standard insulin therapy in adult outpatients with type 2 diabetes requiring dialysis. The trial registration number is NCT04025775.