Mobile agents programming paradigm is an emerging approach for distributed computing, extremely suitable for mobile systems because of its adaptability to exploit the available resources. Optimization of mobile agents applications and system configuration are relevant above all when we deal handheld devices with limited capabilities. Classical approaches are hard to apply because new kinds of interaction facilities provided by agent platforms, such as cloning and migration, represent an additional software layer that affects the system performances. In this context identification and estimation of performance indexes are necessary to evaluate and foresee system dependability. In fact performance evaluation is exploited to address different issues according to which, different measurements are required. In practice, there is a need for tools and techniques for evaluation of the performances of the adopted mobile agent platforms. Related work proposes a set of performance indexes, slightly different one from the other; and measured using different approaches. We present here a framework that aims at supporting the development of ad-hoc solutions based on measurement agents. Framework architecture, prototype implementation and case studies and preliminary performance figures are presented in the following.