This paper aims to initiate a discussion of what an agent in the context of Smart Grid is. But not as usually done from a computational perspective but rather from an engineer's perspective. This discussion seems to be missing with respect to the following questions periodically occurring when Smart Grid researchers get in touch with agent technology. What is the difference between an optimizer or an Energy Management System and an agent? Why are web-services not enough for a future Smart Grid control system? How are multiagent systems structured? These are only some of the questions we will discuss to arrive at an application-oriented definition of an 'agent', understandable for Smart Grid researchers of various disciplines. Fostering such an interdisciplinary discussion seems to be essential when trying to point out the advantages of control systems based on multiagent technologies.