MSMAccelerator is an interface between WorkQueue (from Doug Thain's group), MSMBuilder, and your favorite MD code (GROMACS, OpenMM, Amber, etc).

MSMAccelerator takes control of your MD runs and manages them in an smart way to rapidly converge simulations. It does this by building MSMs and then analyzing which part(s) of the model are statistically weak. MSMAccelerator is then able to drive convergence by placing new simulations in these weak areas, while not wasting sampling exploring already known space.

Further, MSMAccelerator works in a decentralized manner. You boot a master on one computer, and then start up a bunch of workers on remote machines. This allows for efficient use of distributed resources, potentially including cloud servers.