With the rise of simulation in the pursuit of safe autonomous vehicles (AVs) there is a need to be able to work with scenario data in a way that is unrestrictive in its inputs, allows fast reactive simulation of scenarios and enables powerful insight into the interactions and risks present in the data. We propose Scenario Gym ¹¹https://github.com/driskai/scenario_gym as a universal autonomous driving tool to address these needs. Scenario Gym is an opensource, lightweight simulator that allows fast execution of unconfined, complex scenarios containing a range of road users with the ability to gain rich insight via customised metrics and includes a framework for designing intelligent agents for reactive simulation.