We propose an attentive deskwork support system that quickly delivers required objects to people who work at desks. To meet this goal, we propose methods to understand a user’s request for support. To know the existence of the user request, the system uses the characteristics of the user’s hand and eye movements and detects hand-reaching movements. The system understands the content of the user’s request by integrating sensory and contextual information using a probabilistic model. Finally, the system determines a point of delivery by predicting a user’s hand movement and delivers required objects by using self-moving trays. The experiments are conducted to evaluate the usefulness of the system proposed here.