We consider the feasibility of testing Newtonian gravity at low accelerations using wide binary (WB) stars separated by $\ge 3$ kAU. These systems probe the accelerations at which galaxy rotation curves unexpectedly flatline, possibly due to Modified Newtonian Dynamics (MOND). We conduct Newtonian and MOND simulations of WBs covering a grid of model parameters in the system mass, semi-major axis, eccentricity and orbital plane. We self-consistently include the external field (EF) from the rest of the Galaxy on the Solar neighbourhood using an axisymmetric algorithm. For a given projected separation, WB relative velocities reach larger values in MOND. The excess is ${\approx 20\%}$ adopting its simple interpolating function, as works best with a range of Galactic and extragalactic observations. This causes noticeable MOND effects in accurate observations of ${\approx 500}$ WBs, even without radial velocity measurements. We show that the proposed Theia mission may be able to directly measure the orbital acceleration of Proxima Centauri towards the 13 kAU-distant $α$ Centauri. This requires an astrometric accuracy of $\approx 1 \, μ$as over 5 years. We also consider the long-term orbital stability of WBs with different orbital planes. As each system rotates around the Galaxy, it experiences a time-varying EF because this is directed towards the Galactic Centre. We demonstrate approximate conservation of the angular momentum component along this direction, a consequence of the WB orbit adiabatically adjusting to the much slower Galactic orbit. WBs with very little angular momentum in this direction are less stable over Gyr periods. This novel direction-dependent effect might allow for further tests of MOND.

31 pages, 20 figures, 6 tables. Fixed typo in equation 11