All living beings face the universal challenge of foraging (i.e. finding and consuming energy sources while balancing overall energy expenditure). Impressively, in precocial animals such as larval zebrafish, evolutionarily embedded innate neural programs are able to overcome this challenge with no parental guidance and limited learning ability. Previous studies have highlighted the internally driven spontaneous alternation between long-lasting exploration or exploitation states (Flavell et al., 2013; Martin, Ernst, & Heisenberg, 1999, Marques et al., 2020). We aim to understand how these internal states are additionally modulated by the history of the external environment. To this end, we developed a microfluidic system allowing for 1) temporally and spatially precise control of food density in the environment, 2) large field of view behavioral imaging with high spatial and temporal resolution, and 3) brain-wide cellular resolution neural imaging. Within seconds of initial encounter with a new food source (e.g. either live prey or food-associated chemosensory stimuli), we find that larval zebrafish behaviorally adapt by reducing locomotion and entering the exploitation state. Conversely, the removal of food sources induces an exploratory state lasting for several minutes. We identify opposing neural networks representing the presence and absence, as well as the introduction and removal of food in the environment. These behavioral and neural responses to food availability might represent an essential component of an innate foraging strategy in the face of changing environments.