Globally, amphibian species have suffered drastic population declines over the past 40 years. Hundreds of species are now listed as Critically Endangered, with many of these considered “possibly extinct”. Most of these species are stream-dwelling frogs inhabiting remote, montane areas, where traditional surveys limit our ability to find remnant populations. Environmental DNA (eDNA) monitoring could revolutionize surveys for missing and endangered amphibian populations by enabling sampling of downstream sections to screen the vast upstream catchments. However, this is dependent on quantifying downstream detection probability and distances. Here we tested this in two endangered stream frogs (Litoria nannotis and L. lorica) that co-occur in a remote stream catchment, and for which we know precise downstream distributional limits from traditional surveys. Importantly, the last populations of L. lorica persist in this catchment; one small (~1,000 frogs) and one very small (~100 frogs). We conducted eDNA screening at set distances downstream from the populations (up to 22.8 km) using precipitation from two water volumes and via filtering, during moderately high and low flow conditions. During high flow, we reliably detected both species as far as 22.8 km downstream using the larger water volumes and filtering techniques. We could only detect the very small population of L. lorica immediately downstream. Detection success was higher at the sites further downstream for both species during high flow compared to low flow. Our downstream detection distances (> 20 km) suggest eDNA is a valuable tool for detecting rare stream amphibians. We provide recommendations on optimal survey techniques.