We designed a flexible protocol for monitoring fish species richness in nonwadeable rivers. Nine sites were sampled seasonally with six gears in two physiographic regions in Missouri (USA). Using resampling procedures and mixed-effects modeling, we quantified richness and compositional overlap among gears, identified efficient gear combinations, and evaluated protocol performance across regions and seasons. We detected 25–75 species per sample and 89 185 fish. On average, no single gear detected >62% of observed species, but an optimized, integrated-gear protocol with four complementary gears on average detected 90% of species while only requiring 51.9% of initial sampling effort. Neither season nor physiographic region explained low spatiotemporal variation in percent richness detected by the integrated-gear protocol. In contrast, equivalent effort with an electrofishing-only protocol was 53.5% less efficient, seasonally biased and imprecise (36.1%–82.3% of richness), and on average detected 15.9% less of observed richness. Altogether, riverine fish richness is likely underestimated with single-gear survey designs. When paired with existing wadeable-stream inventories, our customizable approach could benefit regional monitoring by comprehensively documenting riverine contributions to riverscape biodiversity.