- Modular multi-agent environment support
- Plugin interface for custom RL algorithms
- Integration with PyTorch and TensorFlow
- Distributed training and GPU acceleration
- Built-in logging, visualization, and metrics
- Scenario configuration and reproducibility tools