Despite the clinical success of pancreatic islet transplantation, graft function is frequently lost over time due to islet dispersion, lack of neovascularization, and loss of physiological architecture. To address these problems, islet encapsulation strategies including scaffolds and devices have been developed, which produced encouraging results in preclinical models. However, islet loss from such architectures could represent a significant limitation to clinical use. Here, we developed and characterized a novel islet encapsulation silicon device, the NanoGland, to overcome islet loss, while providing a physiological-like environment for long-term islet viability and revascularization. NanoGlands, microfabricated with a channel size ranging from 3.6 nm to 60 μm, were mathematically modeled to predict the kinetics of the response of encapsulated islets to glucose stimuli, based on different channel sizes, and to rationally select membranes for further testing. The model was validated in vitro using static and perifusion testing, during which insulin secretion and functionality were demonstrated for over 30-days. In vitro testing also showed 70-83% enhanced islet retention as compared to porous scaffolds, here simulated through a 200 μm channel membrane. Finally, evidence of in vivo viability of human islets subcutaneously transplanted within NanoGlands was shown in mice for over 120 days. In this context, mouse endothelial cell infiltration suggesting neovascularization from the host were identified in the retrieved grafts. The NanoGland represents a novel, promising approach for the autotransplantation of human islets.