Neuropeptides are signaling molecules that commonly act via G protein-coupled receptors (GPCRs) and are generated in neurons by proneuropeptide (pNP) cleavage. Present in both cnidarians and bilaterians, neuropeptides represent an ancient and widespread mode of neuronal communication. Due to the inherent difficulties of analyzing highly diverse and repetitive pNPs, the relationships among different families are often elusive. Using similarity-based clustering and sensitive similarity searches, I obtained a global view of metazoan pNP diversity and evolution. Clustering revealed a large and diffuse network of sequences connected by significant sequence similarity encompassing one-quarter of all families. pNPs belonging to this cluster were also identified in the early-branching neuronless animal Trichoplax adhaerens. Clustering of neuropeptide GPCRs identified several orthology groups and allowed the reconstruction of the phyletic distribution of receptor families. GPCR phyletic distribution closely paralleled that of pNPs, indicating extensive conservation and long-term coevolution of receptor-ligand pairs. Receptor orthology and intermediate sequences also revealed the homology of pNPs so far considered unrelated, including allatotropin and orexin. These findings, together with the identification of deuterostome achatin and luqin and protostome opioid pNPs, extended the neuropeptide complement of the urbilaterian. Several pNPs were also identified from the hemichordate Saccoglossus kowalevskii and the cephalochordate Branchiostoma floridae, elucidating pNP evolution in deuterostomes. Receptor-ligand conservation also allowed ligand predictions for many uncharacterized GPCRs from nonmodel species. The reconstruction of the neuropeptide-signaling repertoire at deep nodes of the animal phylogeny allowed the formulation of a testable scenario of the evolution of animal neuroendocrine systems.