A tool/algorithm for calculating the largest common subgraph between sets of molecules?

Be aware that finding the MCSS between two structures is a different problem from finding them in a larger set. The set MCSS is not necessarily the same, or a substructure of, as any structure pair MCSS from the set. There are papers/published algorithms which get this wrong, and assume that you can start with a structure pair MCSS, and then use that iteratively to shave off more atoms by comparing it to the other structures. That is a faulty approach (though, due to the nature of chemical structures, it still is quite often successful).