Factor XIIa (FXIIa) is a promising target for developing new drugs that prevent thrombosis without causing bleeding complications. A native cyclotide (MCoTI-II) is gaining interest for engineering FXIIa-targeted anticoagulants as this peptide inhibits FXIIa but not other coagulation proteases. Here, we engineered the native biosynthetic cyclization loop of MCoTI-II (loop 6) to generate improved FXIIa inhibitors. Decreasing the loop length led to gains in potency up to 7.7-fold, with the most potent variant having five residues in loop 6 (Ki = 25 nM). We subsequently examined sequence changes within loop 6 and an adjacent loop, with substitutions at P4 and P2' producing a potent FXIIa inhibitor (Ki = 2 nM) that displayed more than 700-fold selectivity, was stable in human serum, and blocked the intrinsic coagulation pathway in human plasma. These findings demonstrate that engineering the biosynthetic cyclization loop can generate improved cyclotide variants, expanding their potential for drug discovery.