What is the latent part of the model? If the mediation part is latent and the moderation variable is binary/categorical, then a multi-group approach may do the job.
Otherwise, if you have to use model indirect then maybe you could get at the indirect effects by using model constraints - label the paths and then create product terms yourself.