Various classes of nanotheranostics have been developed for enhanced tumor imaging and therapy. However, key limitations for a successful use of nanotheranostics include their targeting specificity with limited off-site tissue accumulation as well as their distribution and prolonged retention throughout the entire tumor. Due to their inherent tumor-tropic properties, the use of mesenchymal stem cells (MSCs) as a "Trojan horse" has recently been proposed to deliver nanotheranostics more effectively. This review discusses the current status of "cellular nanotheranostics" for combined (multimodal) imaging and therapy in preclinical cancer models. Emphasis is placed on the limited knowledge of the signaling pathways and molecular mechanisms of MSC tumor-tropism, and how such information may be exploited to engineer MSCs in order to further improve tumor homing and nanotheranostic delivery using image-guided procedures.