I’m using MLForecast and slowly getting a hang of it. However, there seems to be no support for splitting exogenous features into historic and future as there is in Neural Forecast. Is this correct? If so, am I right in understanding that all exogenous features included in the dataframe and not explicitly stated as static_features via the kwarg, are considered future exogenous features?

I believe this is correct and was wondering if it was possible to have historical exo features too. I'm not sure if the algos allow for it. Thanks for asking the question and I look forward to what the folks at Nixtla say.

Yes, that's correct. Some models in neuralforecast support that but machine learning models in general require the same features to be used for training and inference