Using deep ($11.2\, \rm{h}$) VLT/MUSE data from the MEGAFLOW survey, we report the first detection of extended $\rm{Mg\, \small {II}}$ emission from a galaxy’s halo that is probed by a quasar sightline. The $\rm{Mg\, \small {II}}\, \lambda \lambda \, 2796, 2803$ emission around the z = 0.702 galaxy ($\log (M_*/\rm{M_\odot })=10.05_{-0.11}^{+0.15}{}$) is detected out to $\approx 25\, \hbox{kpc}$ from the central galaxy and covers $1.0\times 10^3\, \hbox{kpc}^2$ above a surface brightness of $14\times 10^{-19}\, \rm{erg}\, \rm{s}^{-1}\, \rm{cm}^{-2}\, \rm{arcsec}^{-2}{}$ ($2\, \sigma$; integrated over $1200\, \rm{km\, s}^{-1}= 19\mathring{\rm A}$ and averaged over $1.5\, \rm{arcsec}^{2}$). The $\rm{Mg\, \small {II}}$ emission around this highly inclined galaxy (i ≃ 75 deg) is strongest along the galaxy’s projected minor axis, consistent with the $\rm{Mg\, \small {II}}$ gas having been ejected from the galaxy into a bi-conical structure. The quasar sightline, which is aligned with the galaxy’s minor axis, shows strong $\rm{Mg\, \small {II}}$ absorption ($\hbox{$EW_0^{\lambda 2796}$}{}=1.8{}\, \mathring{\rm A}$) at an impact parameter of $39{}\, \hbox{kpc}$ from the galaxy. Comparing the kinematics of both the emission and the absorption − probed with VLT/UVES − to the expectation from a simple toy model of a bi-conical outflow, we find good consistency when assuming a relatively slow outflow ($v_\rm{out}=130{}\, \rm{km\, s}^{-1}$). We investigate potential origins of the extended $\rm{Mg\, \small {II}}$ emission using simple toy models. With continuum scattering models we encounter serious difficulties in explaining the luminosity of the $\rm{Mg\, \small {II}}$ halo and in reconciling density estimates from emission and absorption. Instead, we find that shocks might be a more viable source to power the extended $\rm{Mg\, \small {II}}$ (and non-resonant $[\rm{O\, \small {II}}]$) emission.