Mean field games with a major agent study optimal control problems with infinitely many small controllers facing a major controller. The "value function" of the agents then satisfy a nonlinear nonlocal system of partial differential equations stated in the space of measures. In this joint work with Marco Cirant (U. Padova) and A. Porretta (U. Rome Tor Vergata) we explain how to build short time a classical solution for this system and use the solution to prove the mean field limit of the associated N player game as the number N of the players tends to infinity.