Cognitive load is the amount of mental resources required to execute a task. Psychophysiological researches show variations in the electrodermal activity of the skin with variations in cognitive load. State of the art studies mostly use exosomatic approach like Galvanic Skin Resistance (GSR) for measuring electrodermal activity. However, recent advances made in implementation of wearable endosomatic devices motivated us to use endosomatic bio-potentials measured using a wearable device for assessing cognitive load. In the present study, we have studied the effect of cognitive load on the endosomatic human bio-potentials. We have used a 4-channel high-resolution wearable device for measuring the bio-potentials during the execution of low and high cognitive load versions of mental addition tasks. Results suggest that there is significant variation in endosomatic electrodermal activity for various levels of cognitive load which can be used to classify low and high mental workload successfully. We have observed and analyzed bio-potentials recorded from both hands and feet. Combining potentials measured from all 4 locations, the classification accuracy obtained using a random forest classifier is approximately 83.4% and the corresponding f-score is 0.83. An optimum set of features selected increases the accuracy further to 86.5%. Moreover, a feature level fusion of GSR and endosomatic signal features gives an accuracy of 89.4%. Thus, the proposed approach provides immense scope for assessing mental workload based on bio-potentials recorded using a wearable device.