Glycosylation patterns commonly change in cancer, resulting in expression of tumor-associated carbohydrate antigens (TACA). While promising, currently available anti-glycan antibodies are not useful for clinical cancer therapy. Here, we show that potent anti-glycan antibodies can be engineered to acquire cancer therapeutic efficacy. We designed yeast surface display to generate and select for therapeutic antibodies against the TACA SLea (CA19−9) in colon and pancreatic cancers. Elite clones showed increased affinity, better specificity, improved binding of human pancreatic and colon cancer cell lines, and increased complement-dependent therapeutic efficacy. Molecular modeling explained the structural basis for improved antibody functionality at the molecular level. These new tools of directed molecular evolution and selection for effective anti-glycan antibodies, provide insights into the mechanisms of cancer therapy targeting glycosylation, and provide major methodological advances that are likely to open up innovative avenues of research in the field of cancer theranostics.