The CityLearn Workshop will introduce the software tool CityLearn to model and analyze advanced control approaches in grid-interactive smart communities, e.g., demand response and load shaping in buildings. CityLearn is an open source OpenAI Gym environment targeted at the easy implementation and benchmarking of advanced control algorithms, i.e., model predictive control or deep reinforcement learning. Main applications to-date consist of controlling the charging and discharging of active storage systems i.e. battery and thermal storage tanks and heat pump power in the buildings to reduce electricity consumption, electricity cost, peak load and carbon emissions.The workshop will be in three parts. The first part will consist of an overview presentation of CityLearn. In the second part, we provide a walk-through tutorial on how to set up the environment using input data from residential building energy models in public End-Use Load Profiles (EULP) for the U.S. Building Stock database. Participants will be able to follow along using the provided Jupyter notebook. The notebook will provide a guide on how to use a simple rule-based control architecture, advanced soft-actor-critic methods, and the MARLISA multi-agent reinforcement learning control architecture. Finally, in the last part participants can optimize hyperparameters of algorithms and compare their findings against each other.CityLearn has been used in 2020--2022 editions of the CityLearn Challenge. The most recent CityLearn Challenge 2022 was hosted on the AICrowd platform (https://www.aicrowd.com/challenges/neurips-2022-citylearn-challenge). It has seen over 500 participants from 50+ countries and 1,500+ submission.