Complementary methods may be used to solve the neutron transport problem. When only a small amount of information is needed, the most efficient method is obviously Monte Carlo. However, when perfect knowledge of the full phase-space is required, it is worth using a deterministic technique. Nevertheless, this memory and CPU time intensive approach may soon overwhelm even the most powerful computer. To deal with these issues, an adapted mesh refinement transport scheme was developed that solely retains active areas of a geometry. The computer code Styx, built on this efficient set of numerical methods, specially designed and tuned to run on such a tree-based topology, is presented. A test case subset, representative of the wide spectrum of multidimensional applications it covers, is then analyzed.