There is no special software needed. Your MAG is either there after the assembly, or it isn't.

A disclosure: I have a bias against binners that require abundance, which is most likely because for years I have been using a homemade program that doesn't need it. Still, binners that need abundance use that information only to establish a consistency between contigs. They don't use that information to eliminate contigs of low abundance.

If you are still worried about it, replace real abundance numbers with a uniform value (say, 10) for all contigs. That way all contigs are assigned equal abundance and that information will not be deciding anything. Instead, binning will be done only by tetranucleotide frequencies.

Finally, a concept of low abundance has different meaning to different people, and it is also relative to the community diversity. In a community with 20 members, one can consider 2% abundance to be low. In a community with 150 members, a 2% abundance would be among dominant members.

In my experience, recovering low-abundance members comes down to the quality of extracted DNA and the subsequent sequencing. In one of the communities I recently analyzed there are ~135 MAGs of which ~100 have abundance < 1%. Of those, ~70 MAGs have completeness > 80%. I have two MAGs with < 0.1% abundance that are 100% complete and < 5% contaminated, and at least 10 MAGs that are abundant at ~0.2% with >95% completeness. I also have evidence beyond this dataset to suggest that the quality of MAG reconstruction is not primarily determined by their abundance.