Considering that neither captology nor agent-orientation, are applied in medical informatics, as they could be, the paper presents a broad-spectrum gene- ric architectural framework to support developing adaptive medical applications, based on synergistic correlation between persuasive interfaces and intelligent agents. Their main features are adapted for medical informatics. Lying on this groundwork, the design space for agent-oriented persuasive applications is defined and several guidelines for its main dimensions are given. The approach is instantiated through an agent-based test-bench application, having the purpose to persuade to quit smoking.