Currently available condition monitoring systems (CMS) offer many types of tools, including stationary systems, portable on-site instrumentation and finally wireless, autonomous systems. However, for a particular group of modern machine tools with limited access, like lathes, milling or grinding machines, none of these tools could be used due to limited access, space constraints, and cabling restrictions. Firstly, stationary CMS use wires connecting vibration sensors with data acquisition unit (DAQ), which are prohibited by both, safety and topological reasons because of a sealed door. Secondly, portable systems could not be used, because modern mining machinery is entirely closed, and a diagnostic engineer is not allowed to be present inside the machine housing while operating. Moreover, some of the machines perform open lubrication, which makes supervised monitoring impossible due to constant lubricant splashes. Tertiary, currently available leading wireless sensors are characterized by relatively significant size, ca. 1.5 by 4 inch, and close to 10 oz. weight; therefore, introducing significant volume and mass to the machine spindle. Moreover, the rapid characteristics of the spindle movement would generate significant inertia to the sensor causing extra vibrations and possible detriment to sensor mounting. For these reasons, modern machine tools are generally not equipped with external CMS systems. Recently developed MEMS technology made it possible to design a novel, small-size unit, which is capable to work autonomously in environments with limited accessibility, where volume and weight matter. In contrast to commonly used piezoelectric accelerometers, MEMS vibration sensors need much less operational power, which is a major concern in wireless designs. Moreover, the latest MEMS sensors are characterized by comparable frequency response to accelerometers. The short-range wireless communication can be applied to avoid connecting cabling. The paper shows a prototype of a micro-size unit for data acquisition, data processing, data storage, and transfer. The prototype is evaluated on the industrial lathe.