In this paper, we describe AUTO-COLLEAGUE that is an adaptive and collaborative learning environment for UML. There are two kinds of user: the trainer and the trainee. The trainees can design UML diagrams given by the trainer or run certain tests/exercises for specific problems that are given. They can collaborate with each other. The system offers them adaptive help and advice regarding the most effective and productive organization of the trainees in groups. This is done so that the users’ collaboration among them can be more efficient. The adaptivity is achieved through user modelling based on stereotypes. The stereotypes are related to three aspects of the performance of the trainee: the Level of Expertise, the Performance Types and the Personality.