Botulinum neurotoxins (BoNTs) deliver a protease to neurons which can cause a flaccid paralysis called botulism. Development of botulism antidotes will require neuronal delivery of agents that inhibit or destroy the BoNT protease. Here, we investigated the potential of engineering Clostridium difficile toxin B (TcdB) as a neuronal delivery vehicle by testing two recombinant TcdB chimeras. For AGT-TcdB chimera, an alkyltransferase (AGT) was appended to the N-terminal glucosyltransferase (GT) of TcdB. Recombinant AGT-TcdB had alkyltransferase activity, and the chimera was nearly as toxic to Vero cells as wild-type TcdB, suggesting efficient cytosolic delivery of the AGT/GT fusion. For AGT-TcdB-BoNT/A-Hc, the receptor-binding domain (RBD) of TcdB was replaced by the equivalent RBD from BoNT/A (BoNT/A-Hc). AGT-TcdB-BoNT/A-Hc was >25-fold more toxic to neuronal cells and >25-fold less toxic to Vero cells than AGT-TcdB. Thus, TcdB can be engineered for cytosolic delivery of biomolecules and improved targeting of neuronal cells.