How does the brain record individual experiences, and recall them to adaptively guide future behavior? This ‘episodic-like’ memory regime involves the hippocampus, but it has been difficult to identify specific neural signatures with recollection of specific memories, hindering a mechanistic understanding of recall. Studying food caching behaviors may provide a clearer picture. Scatter hoarding species, such as chickadees, cache individual food items, and use hippocampal-dependent memory to later retrieve them. However tools for behavioral and neural measurement in chickadees are underdeveloped, and so neural correlates of cache memories have not been identified. We designed a behavioral arena where chickadees spontaneously cache and retrieve seeds, without explicit training or reinforcement. In order to parse this unconstrained behavior, we collect multi-view, high-speed, video data. We combined recent advances in 3D postural tracking with conventional 2D tracking networks into a 2-stream algorithm, which achieved high accuracy with only moderate computational demands. Postural tracking enabled automated detection of multiple discrete behaviors, including interaction with cache sites and gaze location. Finally we developed a light-weight implant for chronic silicon probe recordings compatible with freely moving behavior in small birds, enabling simultaneous recording of ~100 hippocampal neurons. Hippocampal activity during caching exhibited strong modulation preceding cache site interaction. This activity was distinct for each cache site, encoded the hidden contents of the site, and was dissociable from a representation of spatial or other behavioral variables. This activity is consistent with the reactivation of hippocampal codes underlying the recollection of individual caching events.