Selective binding of TCR-based therapies that target a single tumour-specific peptide epitope presented by human leukocyte antigens (HLA) is the absolute prerequisite for their therapeutic suitability and patient safety. To date, selectivity assessment has been limited to peptide library screening and predictive computational modeling. We developed the first experimental platform to de novo identify interactomes of TCR-like molecules directly in human tissues using mass spectrometry. As proof of concept, we confirm the target epitope of a novel MAGE-A4-specific TCR-like antibody. We further determine 16 cross-reactive sequences for ESK1, a TCR-like bispecific antibody recognizing WT-1 with known off-target activity, in healthy liver tissue. We observe strong, off-target-induced T cell activation for 8/16 sequences, demonstrating the high specificity of the approach. Off-target sequences define a previously missed amino acid compensation motif that structurally mimics the target peptide groove coordination and allows for peptide interaction with the engager molecule. We establish the importance of the identified off-target activity by demonstrating 3D liver spheroid killing in the presence of ESK1 and healthy donor PBMC. Finally, we utilize our approach to de-risk our novel MAGE-A4 targeting TCR-like bispecific antibody prior to entering now ongoing clinical trials. We conclude that our strategy offers an accurate, scalable route for de-risking TCR-based therapeutics prior to first-in-human clinical application.