The material accreting onto Sgr A* most probably comes from the nearby stars. We analyze the pattern of this flow at distances of a fraction of a parsec and we argue that the net angular momentum of this material is low but non-negligible, and the initially supersonic disk accretion changes into subsonic flow with constant angular momentum. Next we estimate the flow parameters at a distance R_BHL from the black hole and we argue that for the plausible parameter range the accretion flow is non-stationary. The inflow becomes supersonic at distance of 104 Rg but the solution does not continue below the horizon and the material piles up forming a torus, or a ring, at a distance of a few up to tens of Schwarzchild radii. Such a torus is known to be unstable and may explain strong variability of the flow in Sgr A*. Our considerations show that the temporary formation of such a torus seems to be unavoidable. Our best fitting model predicts a rather large accretion rate of around 4 * 10-6 M_o/ yr directly on Sgr A*. We argue that magnetic fields in the flow are tangled and this allows our model to overcome the disagreement with the Faraday rotation limits.