In external beam radiation therapy, use of sheet bolus can lead to compromises in set-up reproducibility and surface conformity, especially in areas of anatomical complexity. Patient-specific, flexible bolus has been shown to improve those aspects; however, manual fabrication involves using molds, which requires 3D modeling expertise and relies heavily on the skills of the technician. The aim of this work is to demonstrate that using 3D printed 2-piece molds, designed with an automated software algorithm, we were able to easily produce flexible silicone bolus for a variety of geometries, with accuracy in bolus thickness comparable to generic flexible bolus sheets, but with superior surface conformity. An anthropomorphic and a solid water phantom were CT scanned. TPS tools were used to create bolus structures for four anatomical sites (scalp, face, ear and breast) and one slab bolus. DICOM CT and contour data was then imported into the mold generating software. The automated algorithm for design of 2-piece mold STL files was then executed: mold design type was selected; a pouring hole was cut in the resulting mold structure and alignment guides were placed. Both block and shell mold designs were produced for each anatomical site. Mold STL files for 3D printing were exported, along with DICOM RT structure set files, for TPS-based verification of the resulting bolus structure accuracy. Once molds were 3D-printed, silicone rubber was poured. The thicknesses of the patient-specific bolus and a set of generic bolus sheets were measured with an ultrasonic thickness gauge. Patient-specific silicone boluses were positioned on phantoms and CT scanned; this process was repeated with standard bolus sheets taped to cover the same regions. Surface conformity was evaluated and compared. Silicone bolus HU homogeneity was evaluated from the CT data. Mold design took approximately five minutes, with an additional five minutes for TPS-based verification of the resulting bolus structure. 63% of all sheet bolus thickness measurements were found to fall within +/- 0.0 - 0.5 mm of the expected 5 mm and 10 mm sheet thickness and this was taken as the criterion for 3D printed mold-based bolus comparison. All silicone mold-based bolus thickness measurements fell within this criterion, for all anatomic and slab bolus sites and CT images demonstrate superior surface conformity for all anatomical site bolus. Mean silicone bolus density was 1.05 ± 0.04 g/cc and HU uniformity was 150 ± 19 HU. The automated algorithm for designing 2-piece molds is shown to be a simple and versatile method for producing spatially accurate, flexible, patient-specific bolus. 3D-printed mold-based silicone bolus provide superior patient fit and comparable thickness variation when compared to sheet bolus, and also demonstrate clinically acceptable radiological properties.