In this work, a reservoir computing (RC) system implemented with an organic memcapacitor is presented, tailored for energy-efficient time-series classification. The neuromorphic properties of the memory device are shown, and by exploiting them we demonstrate its suitability as a physical reservoir. As a case of study, we implement a system for classification of electrical penetration graph (EPG) signals, which encode the feeding behavior of insects on plant tissues—a critical measure in agricultural pest monitoring. For this task, an accuracy over 93 % is obtained for a reservoir containing eight devices, with an average energy per pulse of 82 pJ per device, showcasing both its high performance and low energy requirement.