The social aspect of today’s web 2.0 applications can play a major role in searching for information on the web. Similar users may share common interests, play same roles, be in the same context or perform the same activities. Search by context has been used as users tend to be looking for the same kind of information. In this work, we complement the search by context by what we call search by role involving the search by particularly and by totally similar users. The model is based on a Multi-Agent System using magnetic users’ cubes where a user’s cube has many facets, or roles, and the search is performed based on these different roles. Users’ related resources, activities, contexts, tags, and relations are collected around each role the user plays.