Bacterial endotoxins are the major mediator of septic shock; therefore, endotoxin-neutralizing molecules could have biomedical applications. The septic shock cascade relies in a series of molecular recognition processes. The large contact-surface described for the interacting macromolecules, in most cases, prevents the identification of small molecules that could modulate such recognition events. Here we report on a beta-hairpin conformationally restricted combinatorial library that has been generated and screened towards the identification of new peptides that neutralize bacterial endotoxins. Starting with a de novo designed linear peptide that shows a beta-hairpin structure population of around 30%, (Ramirez-Alvarado, M., Blanco, F. J. and Serrano, L. Nat. Struc. Biol., 7, 604-612 (1996)), we selected four positions to build up a combinatorial library of 20(4) sequences. Deconvolution of the library reduced such a sequence complexity to 8 defined sequences. The newly identified peptides have a biological activity equivalent to that reported for peptides derived from natural endotoxin-binding proteins.