Service-Oriented Computing allows new applications to be developed by using and/or combining services offered by different providers. Service discovery and composition are performed aiming to comply with the client’s request in terms of functionality and expected outcome. In this paper we present a framework for dynamic service discovery and composition. This framework is based on goals and tasks as the means to represent the client’s expected outcome and functionality, respectively. The framework encompasses a goal-based service ontology, a set of domain and task ontologies and a supporting service platform with a service matching and composition algorithm. The client informs the platform about the goal to be fulfilled. The platform’s matching algorithm searches in the repository for services that can fulfill the client’s goal. Moreover, the platform gathers client’s contextual information to use as inputs for the services and thus, reduce the need for client interaction. If no single service is able to fulfill the user’s goal, a service composition is then performed.