In this paper, we develop the stochastic inventory routing problem for last mile delivery of medical oxygen for hospitals during crisis. We include stochasticity in the inventory routing problem for both split and unsplit deliveries for different hospitals. During the COVID-19 crisis, the time for rebottling of liquid oxygen in jumbo cylinders and further distribution within city hospitals was one of the main reasons for oxygen shortage in smaller hospitals. This problem was faced by many hospitals in India. Further, the capacity constraints of cylinders, geographical limitations in terms of population density, and transportation infrastructure limitations within rural India were some other reasons. In this entire crisis, the Mumbai (India’s financial capital) model was applauded for the protocol of oxygen management, and training the hospitals to prevent oxygen wastage during the COVID-19 crisis. We develop a new inventory routing problem and specifically illustrate the importance of centralized planning and decentralized storage supply to mitigate crisis in the medical oxygen supply chain. The results depict that availability of oxygen can be enhanced through better replenishment policies. Managerial insights and policy implications are addressed through incorporation of resilience and agility in decision making during the crisis.