Edges are ecologically important environmental features and have been well researched in agricultural and urban landscapes. However, little work has been conducted in flammable ecosystems where spatially and temporally dynamic fire edges are expected to influence important processes such as recolonization of burnt areas and landscape connectivity. We review the literature on fire, fauna, and edge effects to summarize current knowledge of faunal responses to fire edges and identify knowledge gaps. We then develop a conceptual model to predict faunal responses to fire edges and present an agenda for future research. Faunal abundance at fire edges changes over time, but patterns depend on species traits and resource availability. Responses are also influenced by edge architecture (e.g., size and shape), site and landscape context, and spatial scale. However, data are limited and the influence of fire edges on both local abundance and regional distributions of fauna is largely unknown. In our conceptual model, biophysical properties interact with the fire regime (e.g., patchiness, frequency) to influence edge architecture. Edge architecture and species traits influence edge permeability, which is linked to important processes such as movement, resource selection, and species interactions. Predicting the effect of fire edges on fauna is challenging, but important for biodiversity conservation in flammable landscapes. Our conceptual model combines several drivers of faunal fire responses (biophysical properties, regime attributes, species traits) and will therefore lead to improved predictions. Future research is needed to understand fire as an agent of edge creation; the spatio‐temporal flux of fire edges across landscapes; and the effect of fire edges on faunal movement, resource selection, and biotic interactions. To aid the incorporation of new data into our predictive framework, our model has been designed as a Bayesian Network, a statistical tool capable of analyzing complex environmental relationships, dealing with data gaps, and generating testable hypotheses.