During the project MED-GOLD, whose aim is to co-develop pilot climate services for three staple Mediterranean agri-food systems: grape, olive and durum wheat, key challenges emerged in the process of identifying useful climate indicators and actionable definitions of the reliability of climate information. To address such conflicting notion of the reliability of climate information, a participatory workshop was organised with providers (mainly climate scientists) and users of climate data and information (representatives from agri-food companies but also providers of agromet services for farmers) to facilitate an open discussion and find ways of moving forward methodologically and practically towards the development of prototype services. We found that the scientists and users had very different conceptions and interpretations of terms such as skill and reliability of climate information. Furthermore, such disparate understandings created a level of friction between what the scientists understood as scientifically robust and credible climate information and what the users required in terms of saliency of the climate information developed in order to effectively support their decisions. Through an iterative and open discussion, scientists and users agreed the decision making landscape and on a notion of reliability of climate predictions connected to the type of decisions that climate information would support. We will describes the process of developing a common understanding on working definitions of the reliability of climate predictions in the MED-GOLD project and, provide a practical example of the application of this definition to a real case study focused on durum wheat cultivation in Italy.