All JNK isoforms play a specific role in various diseases. The role of the JNK2 isoform has so far received little attention compared to its JNK1 and JNK3 counterparts with JNK3 being a potential target for neurodegenerative diseases and an inhibitor with JNK1 bias being currently investigated in clinical trials. Using an iterative, structure-guided optimization approach starting from a reported reversible binding aminopyrazole-derived scaffold, novel highly potent JNK2/3 selective inhibitors were generated ("ligand-first approach"). These reversible inhibitors were further transformed to covalent inhibitors by attaching an electrophilic warhead moiety, able to address a conserved cysteine side chain present in JNKs. Reversible and covalent inhibitors presented in this study show high JNK2/3 isoform selectivity and activity in cells. The covalently acting lead compound 56d shows good kinetic data with a kinact/KI (JNK2) = 38,200 M-1 s-1 as well as cellular isoform selectivity and a clean kinome profile.