OBJECTIVE: Parkinson's disease (PD) is increasingly conceptualized as a disorder of large-scale brain networks, yet whether and how frequency-specific functional connectivity reorganizes across stages remains poorly understood. In this study, we used high-density electroencephalography (EEG) to characterize cortico-cortical functional connectivity across the clinical spectrum of PD. METHODS: We performed high-density EEG in a cross-sectional cohort of 140 PD patients spanning early, intermediate, and advanced stages and 57 healthy controls. Cortico-cortical functional connectivity was reconstructed in source space across multiple frequency bands and analyzed using network-based statistics combined with machine-learning models to identify stage-dependent network alterations and evaluate their diagnostic and prognostic relevance. RESULTS: We detected 3 distinct large-scale networks showing divergent trajectories across disease stages. An α-band network involving prefrontal and parieto-temporal regions exhibited progressive hypoconnectivity and was associated with cognitive and axial impairment. A β-band sensorimotor network showed progressive hyperconnectivity, paralleling bradykinesia severity. A high-γ network demonstrated increased connectivity in early PD, followed by a progressive connectivity breakdown, and was inversely associated with motor complications. Multiband integration achieved near-perfect discrimination between early PD and healthy controls and robust stratification across disease stages. Band-specific networks also predicted clinical milestones of disease progression, with preserved α connectivity identifying patients at lower risk for cognitive and axial impairment, and stronger high-γ connectivity indicating reduced vulnerability to motor complications. INTERPRETATION: Together, these results identify frequency-specific cortical networks as markers of disease stage and clinical vulnerability and support high-density EEG connectivity as a scalable systems-level biomarker for diagnosis, staging, and risk stratification in PD. ANN NEUROL 2026.