Understanding video content relies on analyzing interactions between objects, which can be modeled using scene graphs. However, occlusions often cause information loss, making tracking optimization crucial for complete representation. The multi-camera approach offers a promising solution by combining multiple viewpoints and linking observations of the same object across cameras. We therefore present Neo4MOT, a multi-camera multi-object tracking (MCMOT) method based on a multi-layer temporal graph data model that structures object trajectories, a CNN for visual feature extraction, and a graph algorithm for feature aggregation and trajectory association. The demonstration showcases its execution on the CAMPUS, EPFL, and PETS09 datasets, comparing different tracking algorithms (SORT, OC-SORT, ByteTrack) and enabling trajectory visualization with Neo4j.