Time-varying graph signal processing generalizes scalar graph signals to multivariate time-series data with an underlying graph structure. Important applications include network neuroscience, social network analysis, and sensor processing. In this chapter, we present a framework for modeling the underlying graphs of these multivariate signals along with a filter design methodology based on invariance to the graph-shift operator. Importantly, these approaches apply to directed and undirected graphs. We present three classes of filters for time-varying graph signals, providing example application of each in design of ideal bandpass filters.