Dynamic strains have never played a role in determining local earthquake magnitudes, which are routinely set by displacement waveforms from seismic instrumentation (e.g., M_L⁠). We present a magnitude scale for local earthquakes based on broadband dynamic strain waveforms. This scale is derived from the peak root‐mean‐squared strains (⁠A⁠) in 4589 records of dynamic strain associated with 365 crustal earthquakes and 77 borehole strainmeters along the Pacific‐North American plate boundary on the west coast of the United States and Canada. In this data set, catalog moment magnitudes range from 3.5≤M_w≤7.2⁠, and hypocentral distances range from 6≤R≤500  km. The 1D representation of geometrical spreading and attenuation of A common to all strain data is logA₀(R)=−0.00072R−1.45log(R). After correcting for instrument gain, site terms, and event terms, the magnitude scale, M_DS=log A−log A₀(R)−log(3×10−9)⁠, scales as ≈0.92M_w with a residual standard deviation of 0.19. This close association with M_w holds for events east of the −124° meridian; west of this boundary, however, a constant correction of 0.41 is needed to adjust for additional along‐path attenuation effects. As a check on the accuracy of this magnitude scale, we apply it to dynamic strain records from three strainmeters located in the near field of the 2019 M 6.4 and 7.1 Ridgecrest earthquakes. Results from these six records are in agreement to within 0.5 magnitude units, and five out of six records are in agreement to within 0.34 units.