Autonomic computing is an emergent field for the development of large-scale, self-managing, complex distributed computer-based systems. This paper aims to be a practical approach to Autonomic Computing, by defining and implementing a generic model for autonomic systems, allowing software developers to create autonomic applications using a common and comprehensive infrastructure. In particular, this paper defines a simple but complete model andarchitecture of an autonomic computing element called SelfLet, which could be the building components used to create autonomic systems. SelfLets can be defined by specifying their behaviour, the abilities and goals they need to use and/or provide, and a high-level policy guiding their self-management.