Serine proteases are the most studied class of proteolytic enzymes and a primary target for drug discovery. Despite the large number of inhibitors developed so far, very few make contact with the prime site of the enzyme, which constitutes an almost untapped opportunity for drug design. In the course of our studies on the serine protease NS3/4A of hepatitis C virus (HCV), we found that this enzyme is an excellent example of both the opportunities and the challenges of such design. We had previously reported on two classes of peptide inhibitors of the enzyme: (a) product inhibitors, which include the P(6)-P(1) region of the substrate and derive much of their binding energy from binding of their C-terminal carboxylate in the active site, and (b) decapeptide inhibitors, which span the S(6)-S(4)' subsites of the enzyme, whose P(2)'-P(4)' tripeptide fragment crucially contributes to potency. Here we report on further work, which combined the key binding elements of the two series and led to the development of inhibitors binding exclusively to the prime site of NS3/4A. We prepared a small combinatorial library of tripeptides, capped with a variety of constrained and unconstrained diacids. The SAR was derived from multiple analogues of the initial micromolar lead. Binding of the inhibitor(s) to the enzyme was further characterized by circular dichroism, site-directed mutagenesis, a probe displacement assay, and NMR to unequivocally prove that, according to our design, the bound inhibitor(s) occupies (occupy) the S' subsite and the active site of the protease. In addition, on the basis of the information collected, the tripeptide series was evolved toward reduced peptide character, reduced molecular weight, and higher potency. Beyond their interest as HCV antivirals, these compounds represent the first example of prime site inhibitors of a serine protease. We further suggest that the design of an inhibitor with an analogous binding mode may be possible for other serine proteases.