This paper presents OpenGridGym, an open-source Python-based package that allows for seamless integration of distribution market simulation with state-of-the-art artificial intelligence (AI) decision-making algorithms. We present the architecture and design choice for the proposed framework, elaborate on how users interact with OpenGridGym, and highlight its value by providing multiple cases to demonstrate its use. Four modules are used in any simulation: (1) the physical grid, (2) market mechanisms, (3) a set of trainable agents which interact with the former two modules, and (4) environment module that connects and coordinates the above three. We provide templates for each of those four, but they are easily interchangeable with custom alternatives. Several case studies are presented to illustrate the capability and potential of this toolkit in helping researchers address key design and operational questions in distribution electricity markets.