Finding suitable services to be composed in a service-based business process is always a big challenge. In this paper, we present a query language that addresses this challenge by exploiting relations between services in process models. Con-cretely, we define the composition context of a service as a process fragment around it. We match composition contexts to compute the similarity between services. Finally, we conduct the query based on this computed similarity and flexibly added constraints. Our proposed query helps to retrieve services that have similar composition context with the queried service. It can be applied in process design or execution phase. We perform experiments on a large dataset of real process models. Experimental results show that our approach is feasible and efficient.