The receptor tyrosine kinase EPHA2 plays important roles in oncogenesis, metastasis and treatment resistance but therapeutic targeting, drug discovery or investigation of EPHA2 biology is hampered by the lack of appropriate inhibitors and structural information on how these interact with the protein. Here, we used chemical proteomics to survey 235 clinical kinase inhibitors and identified 24 with sub-micromolar potency for EPHA2. We generated nine high resolution co-crystal structures to identify drug-protein interactions at the atomic level. The combination of compound selectivity, structure determinantion and kinome-wide sequence alignment allowed us to develop a classification system in which amino acids in the drug binding site are categorized into key, scaffold, potency and selectivity residues. This scheme should be generally applicable in kinase drug discovery and we anticipate that the provided information will greatly facilitate the development of selective EPHA2 inhibitors in particular and the repurposing of clinical kinase inhibitors in general.