Graph databases have been developed to meet data persistence requirements, notably from social networks. They are, like the other NoSQL databases, often schemaless. This paper describes an incremental approach deriving a conceptual model from a graph database by analyzing a Cypher flow of queries. This reverse engineering approach embeds three main contributions: (1) a set of transformation rules of Cypher queries into chunks of conceptual schemas, (2) an incremental approach based on these rules, (3) an illustration on an example. This contribution enables, from a Cypher code, to generate a conceptual model that will facilitate the evolution of the existing graph database. This research is part of a project aiming at building an environment enabling round-trip engineering of relational and NoSQL databases.