Railcar air supply systems rely on multiple compressors to maintain critical pneumatic functions such as braking, suspension, and pantograph operation. Currently, compressor operation is typically based on predefined rules and/or single-agent algorithms, which may not be optimal in fluctuating operational conditions, leading to inefficiencies and unbalanced wear. This study proposes the application of Multi-Agent Deep Deterministic Policy Gradient (MADDPG) to dynamically optimize the coordination and operation of these compressors allowing for dynamic and adaptive decision-making, improving energy efficiency, compressor lifespan, and system reliability. The proposed approach leverages individual agents for each compressor to learn coordinated policies based on air pressure, temperature, compressor health, and air demand patterns. Simulation results show significant improvements in energy efficiency, compressor lifespan, and system resilience compared to rule-based strategies.