Localization, personalization, activity recognition, and cognitive assistance are key issues in research on smart homes for cognitively impaired people. Most of the current solutions rely on the presence of solely one person in the residence. To actively consider the interaction of the smart home inhabitant with their caregivers, nurses, doctors and people sharing their home, this paper proposes a multi-agent approach to transparently locate, identify, and ease the collaboration between distributed personalization and assistance services. Based on Bayesian filtering localization using anonymous sensors, the multi-person localization process provides information on each occupant presence, either incoming or outgoing. This information is then used for personalization and assistance.