This article proposes Cultural-Historical Activity Theory (CHAT) as a heuristic model to formalize human-robot interaction in creative contexts. Drawing on CHAT, we discuss the possible alternative roles of the robot in the process of the production of novel and useful artifacts. In contrast to ’disembodied’ computational models of creativity which emphasize abstract reasoning, the physical embodiment of social robots enables the study of how creative artifacts emerge from robots’ actions and bodily engagements with the material and social world. We identify different activity system configurations for human-robot interaction and discuss how the emphasis on action and embodiment allow us to reconsider the structure of the creative process.