The physical and chemical characterization of biological membranes is of fundamental importance for understanding the functional role of lipid bilayers in shaping cells and organelles, steering vesicle trafficking and promoting membrane-protein signaling. Molecular dynamics simulations stand as a powerful tool to probe the properties of membranes at atomistic level. However, the biological membrane is highly complex, and closely mimicking its physiological constitution in silico is not a straightforward task. Here, we present Lipid Builder, a framework for creating and storing models of biologically relevant phospholipid species with acyl tails of heterogeneous composition. LipidBuilder also enables the assembly of these database-stored lipids into realistic bilayers featuring asymmetric distribution on layer leaflets and concentration of given membrane constituents as defined, for example, by lipidomics experiments. The ability of LipidBuilder to assemble robust membrane models was validated by simulating membranes of homogeneous lipid composition for which experimental data are available. Furthermore, taking advantage of the extensive lipid headgroup repertoire, we assembled models of membranes of heterogeneous nature as naturally found in viral (phage PRD1), bacterial (Salmonella enterica, Laurinavicius, S.; Kakela, R.; Somerharju, P.; Bamford, D. H.; Virology 2004, 322, 328-336) and plant (Chlorella kessleri, Rezanka, T.; Podojil, M.; J. Chromatogr. 1989, 463, 397-408) organisms. These realistic membrane models were built using a near-exact lipid composition revealed from analytical chemistry experiments. We suggest LipidBuilder as a useful tool to model biological membranes of near-biological complexity, and as a robust complement to the current efforts to characterize the biophysical properties of biological membrane using molecular simulation.