Sharing data as a (non-)commercial asset on the web is typically performed using an Application Programming Interface (API). Although Linked Data technologies such as RDF and SPARQL enable publishing and accessing data on the web, they do not focus on mediated and controlled web access that data providers are willing to allow. Thus, recent approaches aim at providing traditional REST API layer on top of semantic data sources. In this paper, we propose to take advantage of the new GraphQL framework that, in contrast to the dominant REST API approach, exposes an explicit data model, described in terms of the so-called GraphQL schema, to enable precise retrieving of only required data. We propose a semantic metamodel of the GraphQL Schema. The metamodel is used to enrich the schema of semantic data and enable automatic generation of GraphQL schema. In this context, we present a prototype implementation of our approach and a use case with a real-world dataset, showing how lightly augmenting its ontology to instantiate our metamodel enables automatic GraphQL schema generation.