Conversational agents (CAs) are getting smarter thanks to advances in artificial intelligence, which opens the potential to use them in educational contexts to support (working) students. In addition, CAs are turning toward relationship-oriented virtual companions (e.g., Replika). Synthesizing these trends, we derive the virtual learning companion (VLC), which aims to support working students in their time management and motivation. In addition, we propose design knowledge, which was developed as part of a design science research project. We derive nine design principles, 28 meta-requirements, and 33 categories of design features based on interviews with students and experts, the results of an interdisciplinary workshop, and a user test. We aim to demonstrate how to design VLCs to unfold their potential for individual student support.