This paper presents an agent-oriented approach to modeling software systems for distributed learning to increase the functionality and flexibility of the systems and meet the requirements from the resource-oriented nature of distributed learning environments. We model a distributed learning software system as a multi-agent system using an open collaborative agent system architecture. To overcome the difficulty in deciding what agents we should construct and knowing their relationships, we start the modeling process with identifying the main educational components and the dependency of these components. Meanwhile, in order to facilitate specifying conversation patterns and implementing conversation managers, we examine the workflow of the educational tasks in distributed learning environments. We believe that this approach can be an important step to overcome the software engineering challenges arose in large-scale distributed adaptive learning environments (DALE).