Zebrafish is a widely used animal model in behavioral neuroscience. However, zebrafish learning capabilities are not completely understood. Technological advancements in robotics promise fine behavioral control of artificial conspecifics to study complex aspects of social behavior. In this work, we developed a training system aimed at investigating individual and social learning of zebrafish. The system consists of a shallow water tank, a 2- dimensional robotic platform, and a real-time tracking software. In the tank, a focal individual is separated from a shoal of conspecifics by a one-way glass and a transparent partition, allowing the focal fish to see the shoal. In the transparent partition are two doors, one that automatically opens when the focal individual spends a predetermined amount of time in front of it and another that remains closed regardless of the fish behavior. We tested the system by training one nälve fish in individual learning and one fish in social learning over 20 sessions. Test results show that the fish can learn to open the door and also validated the effectiveness of the developed system applying on individual and social learning.