The purpose of this theoretical essay is to explore the limits of traditional conceptualizations of reader and text and to propose a more general theory based on the concept of a literacy agent. The proposed theoretical perspective subsumes concepts from traditional theory and aims to account for literacy online. The agent-based literacy theory proposed (a) reframes the concepts of reader and text as roles defined by criteria that allow human or machine agents to be assigned to either role, (b) distinguishes two types of literacy events that more adequately represent the range of interactions possible in online reading environments, and (c) argues that readers' views of literacy events and their success in achieving goals depend on how they assign roles in reading. Several examples of literacy events are presented, illustrating the application of agent-based literacy theory and exploring its consequences. Examples are followed by an examination of one historically important critique of the concept of machine readers. The essay concludes by identifying empirical predictions of agent-based literacy theory and reviewing implications of the theory for literacy practice and pedagogy. (Contains 3 figures.)