The prospect of thermal energy harvesting in extreme environments, such as in space or at microscales, offers unique opportunities and challenges for the development of alternate energy conversion technologies. At microscales mechanical friction presents a challenge in the form of energy losses and wear, while presence of high temperature differences and locally available resources inspire the development of new types of heat engines for space and planetary exploration. Recently, levitation using thin-film boiling, via the Leidenfrost effect, has been explored to convert thermal energy to mechanical motion, establishing the basis for novel reduced-friction heat engines. In the Leidenfrost effect, instantaneous thin-film boiling occurs between a droplet and a heated surface, thereby levitating the droplet on its own vapor. This droplet state provides virtually frictionless motion and self propulsion, whose direction can be designed into the system by asymmetrically texturing the substrate. However, sustaining such thermal to mechanical energy conversion is challenging because the Leidenfrost transition temperature for water on a smooth metal surface is similar to 220 degrees C and, despite the low thermal conductivity of the vapor layer, the droplet continuously evaporates. Further challenges include effective transfer of thermal energy into rotational, rather than linear motion, and driving solid components and not simply droplets. Here we present a Leidenfrost rotor, where a solid component is coupled to a rotating liquid volume using surface tension and levitated in continuous operation over a turbine-inspired substrate. We address two key challenges: we show how the liquid can be replenished to achieve the continuous operation of the device; and we show how a superhydrophobic coating to the substrate can broaden the temperature range of operation and the stability of the rotor. Because the liquid acts as a working substance by extracting heat from the substrate to produce useful work in the form of rotation of the coupled solid component, our results demonstrate that a Leidenfrost engine operating in a closed thermodynamic cycle is possible.