To learn more about the topic of  BiFC I can recommend the protocol by Kerppola (doi: 10.1038/nprot.2006.201 ; Design and implementation of bimolecular fluorescence complementation (BiFC) assays for the visualization of protein interactions in living cells ). The paper deals with the in vivo use of BiFC, however it gives a general overview of pitfalls you should consider for in vitro experiments as well.

Do you already have a vector system for cloning or do you want to create the vectors yourself? If so, regarding the structure of the MCS you can check commercial vectors. (This is a vector I used for split YFP in plants (Addgene) and the MCS worked just fine for cloning;  https://www.bio.purdue.edu/people/faculty/gelvin/nsf/protocol_data/BiFC_vectors/pSAT1-nEYFP-C1-seq.pdf )

Do you have any information about the expression behaviour of your proteins in E coli for purification? This is sometimes a problem, when you have proteins which are either toxic or prone to aggregation.

Do you want to check for interaction in general or under specific conditions (e.g. presence of chemicals, altered temperature, etc.)? If you only want to check interaction in general you could also think about analysis in e coli instead of purification.  

If you provide us with more details about your split GFP project we might help. We published a paper with Geoffrey S. Waldo using the system: "Fluorescent Labeling of Antibody Fragments Using Split GFP".

As far as I can tell the Kusabira Green system is basically the same as split GFP, just that the fragmentation is done differently.

I think the setup can work, however you have to control for several factors:

1. Incubation of protein A  (with both tags) without molecule B --> Does your protein A form oligomers?

2. Incubation of only the split GFP fragments in solution (with and without molecule B) --> how high is the background fluorescence from spontaneous reassembly (it is always described that these systems shoudl work "perfectly" without spontaneous reassembly, however I have never seen this. You always get some background fluorescence).

3. One thing I'm not really sure of how it affects the experiment is the mixture of protein A with 2 different tags. In theory you should get equal binding affinities to molecule B for either tagged version. However it might also happen that you have preferrential binding of one molecule species to molecule B which would not result in fluorescence. But this you have to test.

How far are the putative binding sites for the proteins apart on molecule B (how big is molecule B)? If you don't have fluorescence in the final mixture this mightt be due to conformational problems (the two protein A face in different directions, etc.)