Numerous methodologies for developing agent-based systems have been proposed in the literature. However, their application is still limited due to their lack of maturity. Evaluating methodologies’ strengths and weaknesses plays an important role in improving them and in developing the “next-generation” of methodologies. This paper presents a comparison of three prominent agent-oriented methodologies: MaSE, Prometheus and Tropos. It is performed based upon an attribute-based framework which addresses four major areas: concepts, modelling language, process and pragmatics. The objectivity of the comparison is increased by including inputs from the authors of the methodologies using a questionnaire and by conducting an experimental evaluation of the methodologies.