The increasing complexity of the commercial DBMSs as well the workload they manage, besides the fact that many users do not have deep knowledge about database administration, among other reasons, strongly suggests the introduction of techniques that automates the database tuning process. Self- Tuning, or auto-tuning, is a feature that makes systems adaptable in order to keep a good overall performance, reducing as possible the interaction between the administrator and the system. This work proposes an approach for the automatic tuning of DBMSs parameters using an architecture based on software agents. We consider tuning as a global issue, given that changes of a single parameter can be reflected in others. The architecture details, ets implementation and a practical evaluation are also discussed in this dissertation.