We describe a method for the design of multicyclic compounds from three-dimensional (3D) molecular fragments. The 3D building blocks are assembled in a controlled fashion, and closable chains of such fragments are identified. Next, the ring-closing conformations of such formally closable chains are identified, and the 3D model of a cyclic or multicyclic molecule is built. Embedding this method in an evolutionary algorithm results in a de novo design tool capable of altering the number and nature of cycles in species such as transition metal compounds with multidentate ligands in terms of, for example, ligand denticity, type and length of bridges, identity of bridgehead terms, and substitution pattern. An application of the method to the design of multidentate nitrogen-based ligands for Fe(II) spin-crossover (SCO) compounds is presented. The best candidates display multidentate skeletons new to the field of Fe(II) SCO yet resembling ligands deployed in other fields of chemistry, demonstrating the capability of the approach to explore structural variation and to suggest unexpected and realistic molecules, including structures with cycles not found in the building blocks.