Reinforcement Learning (RL) approaches are becoming increasingly popular in the field of combinatorial optimization problems due to their ability to learn and adapt in unknown environments for decision-making problems. In this paper, we propose an RL agent based on the Q-Learning method to address Blocking Job Shop Scheduling (BJSS). This latter can be used to represent many manufacturing and real-world situations by handling storage-capacity constraints. To evaluate the performance of RL, we first model the BJSS problem as an alternative graph and map it to a Markov decision process. Our adapted Q-learning is based on three reward functions and two action selection methods. The results have shown that the Q-Learning method is capable of finding feasible solutions to complex BJSS problems while also learning effective selection strategies for minimizing Makespan. This is the first time a Reinforcement Learning approach has been used to tackle the BJSS problem. It presents a good opportunity for further research and the development of more advanced methods.