The emergence of modular robots has introduced innovative solutions for flexible production lines. These robots can swiftly assemble into various configurations through common interfaces between modules, adapting to diverse tasks. This study focuses on specific tasks, utilizing different joint modules from a modular robot to generate the optimal configuration that meets the task requirements while considering various constraints. We employ a method to automatically generate URDF files for modular robots’ automatic modeling, enhancing the generalizability of our approach. Furthermore, we utilize a two-stage configuration generation scheme to address the issue of excessive computational demands in the configuration generation process. We verified the effectiveness of our scheme by comparing the generated optimal configuration with several lightweight manipulators in a simulation environment, following the completion of a typical task that requires specific trajectories and poses in Cartesian space.