The X-ray crystal structures of the catalytic domain of the EphA3 tyrosine kinase in complex with two type I inhibitors previously discovered in silico (compounds A and B) were used to design type I_1/2 and II inhibitors. Chemical synthesis of about 25 derivatives culminated in the discovery of compounds 11d (type I_1/2), 7b, and 7g (both of type II), which have low-nanomolar affinity for Eph kinases in vitro and a good selectivity profile on a panel of 453 human kinases (395 nonmutant). Surface plasmon resonance measurements show a very slow unbinding rate (1/115 min) for inhibitor 7m. Slow dissociation is consistent with a type II binding mode in which the hydrophobic moiety (trifluoromethyl-benzene) of the inhibitor is deeply buried in a cavity originating from the displacement of the Phe side chain of the so-called DFG motif as observed in the crystal structure of compound 7m. The inhibitor 11d displayed good in vivo efficacy in a human breast cancer xenograft.