The comprehensive optimization of building energy design is fundamental to promote sustainability but it is an arduous issue that involves a huge domain of variables and objectives. The proposed investigation addresses this issue through a novel comprehensive framework - Harlequin - that performs a multi-phase and multi-objective design optimization. Three phases are carried out to optimize design variables related to the whole building-plants system, considering different energy, comfort, economic and environmental performance indicators. Phase 1 implements a genetic algorithm to achieve the Pareto optimization of envelope, geometry and space conditioning set points. Phase 2 performs a smart exhaustive sampling of design scenarios to find optimal energy systems. Phase 3 provides the most sustainable, the cost-optimal and the lowest investment (but energy-efficient) design solutions. Among these, the stakeholders can choose the best solution according to their wills and needs. Harlequin uses EnergyPlus (only in phase 1) and MATLAB (R) and it is so-called because building geometry and envelope are optimized for each exposure, thereby providing "Harlequin buildings". The novelty and scientific significance consist in ensuring a reliable design optimization by investigating a domain of variables and objectives, as comprehensive as never before. As a case study, Harlequin is applied to design a typical Italian office in Milan. Compared to a reference design, significant reductions of primary energy consumption (PEC), global cost (GC) and CO2-eq emissions can be achieved, depending on the chosen solution. The maximum reductions are 43.9 kWh(p)/m(2) a for PEC, 63.9 (sic)/m(2) for GC (discount rate of 3%) and 12.3 kg/m(2) a for CO2-eq.