Developing new AI algorithms may require researchers to design new environments that can facilitate learning. Although several tools exist to help AI researchers customize 3D simulation environments, there are fewer tools available for creating 2D environments. Therefore, to facilitate multi-agent deep RL research and experimentation, DeepMind has released Lab2D, an OSS scalable environment simulator designed to create 2D, layered, discrete grid-world environments. These grid world environments are represented by simple text-based maps that describe the layout of the world and Lua code that define its behavior. ML agents can interact with the environment through a Python or C API.