The primary autoantigen triggering spontaneous type 1 diabetes mellitus in nonobese diabetic (NOD) mice is insulin. The major T-cell insulin epitope lies within the amino acid 9–23 peptide of the β-chain (B:9–23). This peptide can bind within the peptide binding groove of the NOD MHC class II molecule (MHCII), IA^g7, in multiple positions or “registers.” However, the majority of pathogenic CD4 T cells recognize this complex only when the insulin peptide is bound in register 3 (R3). We hypothesized that antibodies reacting specifically with R3 insulin–IA^g7 complexes would inhibit autoimmune diabetes specifically without interfering with recognition of other IA^g7-presented antigens. To test this hypothesis, we generated a monoclonal antibody (mAb287), which selectively binds to B:9–23 and related variants when presented by IA^g7 in R3, but not other registers. The monoclonal antibody blocks binding of IA^g7-B:10–23 R3 tetramers to cognate T cells and inhibits T-cell responses to soluble B:9–23 peptides and NOD islets. However, mAb287 has no effect on recognition of other peptides bound to IA^g7 or other MHCII molecules. Intervention with mAb287, but not irrelevant isotype matched antibody, at either early or late stages of disease development, significantly delayed diabetes onset by inhibiting infiltration by not only insulin-specific CD4 T cells, but also by CD4 and CD8 T cells of other specificities. We propose that peptide–MHC-specific monoclonal antibodies can modulate autoimmune disease without the pleiotropic effects of nonselective reagents and, thus, could be applicable to the treatment of multiple T-cell mediated autoimmune disorders.