To construct a recombinant expression system for a single-domain antibody targeting the urease of Helicobacter pylori (Hp), this study employed several strategies. First, using artificial intelligence (AI) auxiliary tools such as Pymol, I-TASSER, and ClussPro2, the molecular interactions between different antibodies and Hp urease subunit B (UreB) were analyzed. The fully humanized single-domain antibody UreBAb was identified as the primary research target. Next, the UreBAb gene sequence was optimized based on Escherichia coli codon preferences, and was inserted into expression vectors such as pET28a and pE-SUMO. The resulting recombinant expression strains were obtained by transforming Escherichia coli Rosetta(DE3). Recombinant antibody proteins were prepared through IPTG induction, and its activity was detected using extracted Hp urease as the antigen. SDS-PAGE analysis confirmed the correct expression of both UreBAb and SUMO-UreBAb, with protein yields of 0.34 mg/mL and 0.41 mg/mL, respectively. Unidirectional immunodiffusion experiments further confirmed that both recombinant antibodies exhibited strong affinity for Hp UreB antigen, with inhibition rates of 51.27% and 74.07%, respectively. Additionally, leveraging artificial intelligence tools such as AlphaFold2, cluspro2, mCSM-AB, OSPREY, and FoldX, the study evaluated and analyzed key binding sites and mutational strategies affecting the stability of the antigen-antibody complex. Subsequently, nine UreBAb evolution mutants were constructed, and their binding activities with the antigen were enhanced. Among these, the I107W mutant showed the most significant improvement, achieving a 24.95% increase compared to the wild-type UreBAb. This research lays a solid foundation for the development of fully humanized single-domain antibodies against Hp.