How can short-lived molecules selectively maintain potentiation of synapses to sustain long-term potentiation (late-LTP) and long-term memory? One possibility, suggested by Francis Crick (Nature, 1984), is that continual interaction between synaptic molecules maintains the strengthening of activated synapses in the face of molecular turnover. Here, we tested the hypothesis that continual interaction between KIBRA, a postsynaptic scaffolding protein genetically linked to human memory performance, with the autonomously active, atypical PKC isoform PKMζ, maintains late-LTP and long-term memory. Using immunohistochemistry to visualize colocalization of KIBRA with PKMζ, and in situ proximity ligation assay (PLA) to directly detect KIBRA-PKMζ complexes in parallel, we found that strong synaptic stimulation facilitates increases of total KIBRA and PKMζ and formation of persistent KIBRA-PKMζ complexes at synaptic sites. We causally tested whether persistent KIBRA-PKMζ coupling maintains synaptic potentiation and memory, using ζ-stat or K-ZAP, two structurally distinct antagonists of KIBRA-PKMζ dimerization validated with the split-Venus bimolecular fluorescence complementation reporter assay (BiFC). Both antagonists reversed established late-LTP without affecting basal synaptic transmission. ζ-stat or K-ZAP also disrupted the long-term retention of spatial and fear/threat memory. Neither antagonist affected late-LTP or long-term memory compensated by other PKCs in PKMζ-null mice; thus, both ζ-stat and K-ZAP require PKMζ for their effect. KIBRA-PKMζ complexes maintained 1-month-old memory despite PKMζ turnover. Our results reveal that it is not PKMζ alone, nor KIBRA alone, but the continual interaction between the two that maintains late-LTP and long-term memory.