We present a new method for determining the main relevant features of the local magnetic field configuration, based entirely on the knowledge of the magnetic field gradient four‐spacecraft measurements. The method, named “magnetic configuration analysis” (MCA), estimates the spatial scales on which the magnetic field varies locally. While it directly derives from the well‐known magnetic directional derivative and magnetic rotational analysis procedures (Shi et al., 2005, htpps://doi.org/10.1029/2005GL022454; Shen et al., 2007, https://doi.org/10.1029/2005JA011584), MCA was specifically designed to address the actual magnetic field geometry. By applying MCA to multispacecraft data from the Magnetospheric Multiscale (MMS) satellites, we perform both case and statistical analyses of local magnetic field shape and dimensionality at very high cadence and small scales. We apply this technique to different near‐Earth environments and define a classification scheme for the type of configuration observed. While our case studies allow us to benchmark the method with those used in past works, our statistical analysis unveils the typical shape of magnetic configurations and their statistical distributions. We show that small‐scale magnetic configurations are generally elongated, displaying forms of cigar and blade shapes, but occasionally being planar in shape like thin pancakes (mostly inside current sheets). Magnetic configurations, however, rarely show isotropy in their magnetic variance. The planar nature of magnetic configurations and, most importantly, their scale lengths strongly depend on the plasma β parameter. Finally, the most invariant direction is statistically aligned with the electric current, reminiscent of the importance of electromagnetic forces in shaping the local magnetic configuration.