Magnesium phosphate cements (MPCs) are receiving increasing attention because of their outstanding properties of fast setting and hardening, good volume stability and excellent bonding to old concrete structures. Their main area of application is thus rapid repair works, but they may also offer prospects for the stabilization / solidification of deleterious waste.The main constituents of MPCs are magnesium oxide (calcined, or hardburnt, magnesia) and a water-soluble acid phosphate, which is most often diammonium hydrogen phosphate (NH$_4$)$_2$HPO$_4$. To avoid the release of noxious gaseous ammonia during the hardening process, potassium dihydrogen phosphate (KH$_2$PO$_4$) is used. The main precipitated hydrate is then K-struvite (MgKPO$_4$.6H$_2$O). This paper aims at giving new insight into the processes involved in its formation. Since the reaction between magnesium oxide and potassium phosphate is very rapid and needs to be retarded for field application, the second objective of this work is to understand how boric acid, a common admixture, retards the precipitation of K-struvite.MPC was prepared by mixing hard-burnt magnesia and KH$_2$PO$_4$ in equimolar amounts. Cement pastes comprised MPC, low-CaO fly ash used as a filler (fly ash-to-cement weight ratio of 1), and water (water-to-cement weight ratio of 0.56). Additional experiments were performed on cement suspensions (water-to-cement ratios equal to 10 and 100). The boric acid concentration was fixed at 4.17 mmol/L for the study. A panel of techniques was used to investigate the early age hydration of MPC. The evolution of the shear storage modulus, electrical conductivity and heat flow of the cement pastes were monitored with ongoing hydration. In addition, the phase assemblage was characterized after fixed periods of time by X-ray diffraction and thermogravimetry. Complementary experiments on suspensions allowed the evolution of the solution composition to be determined with time.A multi-stage process was evidenced, depending on the w/c ratio precipitation of K-struvite could be preceded by the successive formation of newberyite (MgHPO$_4$.3H$_2$O) and Mg$_2$KH(PO$_4$)2.15H$_2$O. In some cases; Mg$_3$(PO$_4$)$_2$.22H$_2$O precipitated in addition to K-struvite. Experiments on cement suspensions showed that boric acid did not slow down the initial dissolution of magnesium oxide, but rather retarded the precipitation of the products. The total boron concentration in solution remained constant during the whole hydration process. The MgB(OH)$_4$$^+$ complex, stabilizing Mg in solution, might be involved in the retardation process. Moreover, in basic medium, boric acid was dissociated into anionic forms (B(OH)$_4$$^-$, and polyborates at high boron concentration). These negative charges were compensated by an increase in the aqueous concentration of potassium, which in turn tended to favor the formation of Mg$_3$(PO$_4$)$_2$.22H$_2$O against that of K-struvite.