Label-free multiphoton microscopy is a powerful tool for investigating pristine biological specimens. This imaging modality leverages optical signals originating from the nonlinear response of native biomolecules to intense optical radiation, nonlinear signals that allow localizing and quantifying the constituents of specimens, driving applications in biology and medicine. However, since its inception over three decades ago, this approach has operated with a narrowband detection scheme, relying on narrow bandwidths from the entire spectra to derive imaging contrast. This detection scheme hinders the analytical power of the nonlinear microscope, preventing the rigorous unmixing of co-localized constituents with spectral overlap. In this Letter, we shift from the narrowband paradigm to the broadband, demonstrating label-free hyperspectral multiphoton imaging of biological specimens. We validate this configuration by disentangling the constituents of fresh murine tissues by virtue of nonlinear spectra.