Brain networks communicate through anatomical connections which are thought to mediate flexible, state-dependent interactions. We call the resulting pattern of communication “effective” or “functional” connectivity. In the motor system, within-day learning may rely primarily on modulation of networks, although the source of this modulation is unclear [1]. Over longer timescales, changes in anatomical connectivity can shape behavior through processes like Hebbian plasticity [2]. Understanding how these mechanisms combine to shape learning will require measurements of anatomical and functional connections as animals learn. However, in vivo measurements of connectivity that rely on activity are prone to errors in network estimation [3] and it’s unclear how they relate to anatomy. Using stimulation to measure connectivity can improve accuracy and may be a better proxy for anatomical connectivity because it more directly probes pathways from the stimulation site. Recent work also detected trial-to-trial modulation of stimulation-evoked connections in long-range brain networks [4]. Here, we present a novel approach to longitudinally measure stimulation-evoked functional connectivity in cortical networks and experiments to test whether these measurements can capture state-specific modulation. We revealed networks of significant connections using paired optogenetic stimulation and micro-electrocoticogram (µECoG) recordings across a window covering pre-frontal, pre-motor, and primary motor cortices in non-human primates. We repeated the measurements across days at rest and in two behavioral tasks: visual stimuli and goal-directed reaches. Connections were stable across days but increased significantly in both tasks. Because the measured responses were always driven by stimulation, these state-dependent increases suggest we are capturing flexible modulation of connectivity. Our results demonstrate a platform for measuring multi-area interactions chronically which will help us understand how anatomical and functional connectivity changes may contribute to learning over long timescales. These experiments may also refine our ability to model multi-area interactions by leveraging anatomical data in combination with neural activity [5].