Distance learning involves a lot of work of human assistants. These assistants need to be connected for answering student doubts and questions. Intelligent agents can do part of this repetitive work because they can observe students interacting with educational courses, detect learning troubles of these students, and then suggest them some way for overcoming those troubles. However, a design problem appears with this promised possibility: how to connect educational applications with these agents. This paper presents a solution to this problem, in which both the capture of student's intentions and agent intervention for helping students are specified. These two architectural design points are defined as connection points. The first connection point is named student intentions. Student intentions define situations in which agents might help. This connection point depends on the user interface of the educational application that students are using; the agent needs to know the gestures that students could do for interpreting their intentions. The second connection point is named agent interventions. Agent interventions define the context in which agent might assist and the type of help that might give, like a suggestion or a warning. This solution is introduced in the context of one specific application for distance learning named SAVER, which is used for exemplifying each architectural design point.