Clinical magnetic nanoparticle hyperthermia therapy (MNHT) requires controlled energy deposition to achieve a prescribed tumor thermal dose. The objective of this work is to design a thermal dose feedback control to deliver prescribed Cumulative Equivalent Minutes at 43 [°C] (CEM43) based on values at selected tumor boundary points. Constraints were imposed to maintain the maximum treatment temperature below 60 [°C] and the tumor boundary at ∼ 43 [°C]. The controller was designed by performing an integrated system dynamic – finite element analysis. Finite element-bioheat transfer (FE-BHT) simulations were performed on a computational phantom developed from the imaging data of a de-identified human head divided into voxels representing the skull, cerebrospinal fluid (CSF), brain, tumor, and ventricles. A uniform distribution of magnetic nanoparticles (MNPs) in an ellipsoid was used to represent MNPs in the phantom tumor. The MNP distribution was subdivided into three domains to simulate the steerable spatially confined heating region during MNHT. Proportional-integral-derivative (PID) control and model predictive control (MPC) were explored. Regions of the phantom tumor that were undertreated during the simulated MNHT were selectively heated by adjusting the heating volume to improve the tumor coverage index (CI; tumor volume ≥ CEM43 of 20 [min]). Results show that steerable spatially confined heating improves CI by ∼15%. MPC achieves CI of 80% faster than PID (67 [min] vs. 80 [min]). Simulations demonstrated the feasibility of automated control to deliver tumor conformal thermal doses using steerable spatially confined heating.