Graph databases have surged in popularity, and applications increasingly employ them to store and retrieve interconnected data. However, testing graph database-backed applications has distinctive challenges. Due to the sheer dimension of the graph schema state space, testing applications using naive random graph instances is unlikely to cover a large portion of an application program. We present PGFuzz, a graph transformation-based greybox fuzzer for testing graph database-backed applications, that is, to the best of our knowledge, the first fuzzer to specifically target graph database applications. PGFuzz builds on top of state-of-the-art graph generators and utilizes graph transformations guided by code coverage to produce application test inputs. PGFuzz’s graph transformations are schema-aware and support recently introduced graph schema, key, and cardinality constraints. We evaluate PGFuzz on graph database applications that we curate from open-source repositories and show that PGFuzz substantially improves the test coverage of graph database-backed applications compared to the state of the art.