Recently Machine Learning techniques have been applied to the Traveling Salesman Problem to find approximate solutions. Reinforcement Learning in particular is promising as it has achieved superhuman performance. This superior performance is attributed to allowing the model to discover unknown strategies without restricting it to those that are provided as implemented in Supervised Learning approaches. Instead, this paper proposes guiding and supervising, not confining, a Reinforcement Learning approach to solve the TSP with the strategies offered by Search Heuristics.