The goal of cancer treatment is to remove or kill malignant cells while preserving surrounding healthy tissue. Among treatment methods, needle-based ultrasound thermal ablation is an option that involves the insertion of an applicator into the patient's body and the use of an ultrasound transducer to vibrate tissue, producing heat. An ablation pattern for an arbitrarily shaped tumor can be approximated by moving the applicator to deposit heat in targeted locations. However, this conformal ablation process is challenging to control because of the complex interactions between tissue and ultrasound. To address this, we built an interactive planning toolkit that allows a physician to perform the procedure multiple times in simulation and record the ablation trajectory once a desired result is achieved. To validate this method, a previously developed MRconditional robot was used to replicate the planned ablation in a phantom model. Live Magnetic Resonance Thermal Imaging was used to track temperature changes, allowing us to measure the thermal dose and identify the ablated region. In 4 ablations experiments, we achieved an average of 80.9% overlap between the targeted tumor area and the actual ablated area, with minimal damage of 9.4% affected surrounding tissues, demonstrating the effectiveness of our approach.