The Korea Microlensing Telescope Network (KMTNet) Alert Algorithm and Alert System.

We describe a new microlensing-event alert algorithm that is tailored to the Korea Microlensing Telescope Network (KMTNet) multi-observatory system. The algorithm focuses on detecting events, i.e., events whose brightness is increasing as a function of time. The algorithm proceeds in three steps. It first identifies light curves with at least $N_{rm high}$ points that are at least $3sigma$ above the median and that had been taken since shortly before the previous search for new events. It then demands that there are at least $N_{rm high}$ $textit{consecutive}$ high points when considering any combination of light curves from one, two, or three observatories. Finally, it fits to a microlensing model consisting of a broken line, i.e., flat before some time $t_{rm rise}$ and rising linearly afterward. If this fit is better than a flat line by $Deltachi^2u003eDeltachi^2_{rm thresh}$, the light curve is sent for human review. Here, $(N_{rm high},Deltachi^2_{rm thresh})=(5,250)$ or $(10,400)$, depending on the field cadence. For 2018, KMTNet alerts will initially be restricted to a few northern bulge fields and may gradually extend to the full northern bulge. Further expansion of coverage is expected in 2019.