Conversational agents like virtual assistants or chatbots are getting more and more attention within research and practice due to a significant technological progress in the field of Artificial Intelligence (AI). This interdisciplinary field of research led to a variety of different terms, approaches, frameworks, guidelines and views, that in consequence challenges organizations and researchers to systematically understand, design and implement AI-based applications like conversational agents. Moreover, use- cases in which conversational agents do not just assist their user, but rather collaborate with them, will be interesting in the future. We therefore propose the collaborative agent, an evolution of the conversa- tional agent and introduce the Virtual Companion (VC), an exemplary class of the collaborative agent. As profound design theories and tool support for the design of AI is missing, we introduce the Virtual Companion Canvas (VCC), an approach towards a design theory and a tool to analyze, understand and design VCs, detached from concrete technology. The VCC evolved within a comprehensive Design Sci- ence Research project, with currently two contributing design projects containing three design cycles and two successive iterations of the VCC itself. This research paper introduces the concept of the col- laborative agent and reports the development of the VCC as well as the results of a conducted experiment.