Is it better to add control variables before or after the main predictor variable while conducting a step-wise multilevel model?
If you want to see what the control variables do to the main independent variables, then you need to do two runs: One with just the IVs and one with the IVs plus the control variables.
This is often a good idea, as it helps get away from automatic methods of variable selection. It may give you a reason to keep control variables that are very far from significant in the model, as they may affect the parameter estimates and standard errors of the IVs.
This is true regardless of whether it's a regular OLS regression, a MLM, some other kind of regression (e.g. logistic) or whatnot.
It encourages thought on the part of the data analyst. That is a good thing.