As discussed in the accompanying paper [Markland, W., Ley, A. C., & Ladner, R. C. (1996) Biochemistry 35, 8045−8057], we generated libraries from the first Kunitz domain of human lipoprotein-associated coagulation inhibitor (LACI-D1) using multivalent M13 III display and derived potent inhibitors of human plasmin (PLA) by iterative variegation and selection. Here, we show that high-affinity, high-specificity binders to human plasma kallikrein (pKAL) and human thrombin (THBN) can be obtained starting from the identical library and employing the same iterative variegation procedures used to obtain PLA inhibitors. Lib#1 (allowing 31 200 variants involving five positions near the P1 residue of LACI-D1) and its pKAL-biased derivative, Lib#4 (allowing an additional 1600 variants at residues 31, 32, 34, and 39), were screened against pKAL, yielding potent inhibitors. One of these, EPI-K401, has K_i = 284 pM, very high specificity, and excellent stability. We used information from Lib#4 selectants to design Lib#5 (allowing 1.5 × 10⁶ amino-acid sequences involving nine varied positions) from which we obtained an inhibitor (EPI-K503) having high affinity for pKAL (K_i = 40 pM) and retaining the high specificity of EPI-K401. When we screened Lib#1 and its THBN-tailored derivative, Lib#6, against THBN, we obtained a different and very homogeneous population of selected molecules. The purified proteins derived from Lib#6 selectants bound to THBN-agarose beads but did not inhibit proteolytic activity of THBN, suggesting that these selectants bind to a site on THBN other than the catalytic site. Thus, a single large combinatorial library can serve as a source to obtain highly specific, high-affinity binding molecules for each of several targets. Furthermore, the results with THBN show that the binding of Kunitz domains to other proteins is not limited to the catalytic sites of trypsin-homologous proteases.