Data storage in various systems such as SQL and NoSQL leads to important problems when trying to unify data querying. Multiple storage systems conduct to heterogeneous data structures and to multiple query languages. In the context of horizontally and disjointed distributed data, this paper proposes a system that allows the user to natively query a polystore system without taking care of data distribution and heterogeneity. Our approach relies on two mechanisms: (i) mapping dictionaries to define the navigation between systems, (ii) operator rewriting mechanisms from native query operators (selection, projection, aggregation and join) to execute queries on any polystore system. Using a dataset from TPC-H benchmark and a horizontally distributed between document and relational database management system, we conduct experiments showing that the rewriting process has a minimum impact when compared to executing queries in both systems.