Knowledge Graphs are a powerful tool for the organization of knowledge, especially when it is easily represented in textual form. For knowledge best represented using other modalities, such as vision or sound, Multimodal Knowledge Graphs incorporate multimedia documents into the graph, using either BLOB literals or URI references to documents external to the graph. This makes the media documents opaque to the graph query engine and limits accessibility to the information contained within. In this paper, we demonstrate MeGraS, the MediaGraph Store, a storage and query engine for Multimodal Knowledge Graphs. MeGraS stores documents directly and makes them available as graph nodes through RDF-compatible URIs. It further offers facilities to address not only complete documents but also arbitrary spatiotemporal segments as part of the graph and dynamically derive information from them at query time. MeGraS can be queried using SPARQL and supports several custom extensions to enrich the query functionality when working with multimodal data. MeGraS is available as open-source software via https://megras.org/ , and a demonstration video showing MeGraS in action can be seen on https://megras.org/MMM2026demo .