The application would need to be written to be able to cluster. There are several projects that help implement that clustering, but for a typical application that wasn't originally written to cluster it would require a complete re-write.
Only certain computations will benefit from clustering, work that doesn't depend on the output of other work is the best result of clustering. Work that needs the output from another worker before it can begin to compute is the worst case. Eventually even the best case for a program won't scale linearly either; 5 computers crunching won't be five times as fast, they might be 4.5 times or closer to 5, but it won't be exactly 5.
It is not, repeat not possible to take an unmodified program that is not designed specifically to run on multiple machines to take advantage of another computer's resources.