We propose efficient algorithms to compute tight lower bounds and high quality upper bounds for the Multi-Level Capacitated Minimum Spanning Tree problem. We first develop a branch-and-cut algorithm for the problem. This algorithm is able to solve instances of medium size and to provide tight lower bounds for larger ones. We then use the branch-and-cut within GRASP to evaluate subproblems during the search. The computational experiments conducted have improved best known lower and upper bounds on benchmark instances
