High-grade glioma is an incurable brain cancer with a median survival of approximately 14 months. Over the last 50 years, small improvements in patient outcomes have been overshadowed by significant progress in most other cancers. Yet, emerging research has revealed that neural circuits play an active and central role in driving glioma growth and proliferation, highlighting the nervous system as a promising avenue for both disease monitoring and therapeutic intervention. Here, we present a platform for chronically monitoring tumor progression using neural recordings in freely behaving mice with gliomas. Using this platform across multiple mouse strains and glioma models, we show that neural recordings can accurately track tumor progression in vivo. Cancer progression was consistently associated with elevated gamma-band neural activity in the tumor microenvironment across both adult glioblastoma (GBM) and pediatric diffuse intrinsic pontine glioma (DIPG) cancer models. Interestingly, lower frequency neural activity exhibited distinct, cell-line specific changes over time: GBM models exhibited decreases in low frequency neural activity whereas DIPG models exhibited increases over time. Finally, using machine learning models applied to chronic neural recordings from tumor-bearing mice treated with or without standard-of-care chemotherapy (temozolomide for GBM), we accurately predicted tumor burden as inferred through in vivo bioluminescence imaging. By fitting low-dimensional mathematical models to gamma-band neural trajectories, we could further predict individual tumor growth rates over a 5-week period with high accuracy. These results establish that pathological neural–tumor interactions can be harnessed to monitor glioma progression in vivo. Coupling this monitoring capability with therapeutic electrical stimulation in the same device could open up a new class of implantable, closed-loop neurotechnologies with the potential to transform glioma treatment.