The choice in selection, crossover and mutation operators can significantly impact the performance of a genetic algorithm (GA). It is found that the optimal combination of these operators are dependent on the problem characteristics and the size of the problem space. However, existing works disregard the above and focus only on introducing adaptiveness in one operator while having other operators static. With adaptive operator selection (AOS), this paper presents a novel framework for an adaptive and modular genetic algorithm (AMGA) to discover the optimal combination of the operators in each stage of the GA’s life in order to avoid premature convergence. In AMGA, the selection operator changes in an online manner to adapt the selective pressure, while the best performing crossover and mutation operators are inherited by the offspring of each generation. Experimental results demonstrate that our AMGA framework is able to find the optimal combinations of the GA operators for each generation for different instances of the traveling salesman problem (TSP) and outperforms the existing AOS models.