The adoption of a software agent-oriented development approach brings a number of benefits such as the support of analyses, design and implementation of software systems. We posit that a robust development environment for the construction of agent-oriented software systems will be enhanced by advanced reuse methods. Reuse in software engineering leads to the reduction of cost and time involved for developing projects, improvement of software quality, and easy adaptation to different requirements and needs of software engineers. However, research addressing agent reuse is meager and does not tackle the problem of organizing and storing agent-oriented artifacts according to the software engineers needs. Therefore, the agent retrieval process turns into an important challenge to be overcome in agent-oriented software engineering. In this context, this thesis proposes a methodology and a repository prototype based on semantic web technologies that supports reuse for developing agent systems. The proposed methodology includes (i) a flexible meta-model for representing the agents and their common and variable features, which are represented by means of ontology and feature model, (ii) a domain-specific taxonomy to classify agents according their application domain, (iii) a taxonomy to classify the different behaviors of agents, (iv) a subscription service, RSS, to announce updates to the agents that are associated to specific categories, (v) a recommendation system, supported by the ontology and the taxonomies, that allows end-users to discover reusable interrelated agents and learn new information or agents as needed, and (vi) enhanced search and browsing methods for agents. Finally, we evaluate the constructed repository. The outcome of this evaluation shows that the proposed reuse method is an improvement in terms of the relevance of retrieved agent-oriented artifacts.