Stream restorations are increasingly critical for managing and recovering freshwater biodiversity in human-dominated landscapes. However, few studies have quantified how rehabilitative actions promulgate through aquatic communities over decades. Here, a long-term dataset is analyzed for fish assemblage change, incorporating data pre- and post-restoration periods, and testing the extent to which native assemblage stability has increased over time. In the late 1950s, a large capacity dam was installed on Putah Creek (Solano County, CA, USA), which altered the natural flow regime, channel structure, geomorphic processes, and overall ecological function. Notably, downstream flows were reduced (especially during summer months) resulting in an aquatic assemblage dominated by warm-water nonnative species, while endemic native species subsisted at low levels as subordinates. A court-mediated Accord was ratified in 2000, providing a more natural flow regime, specifically for native and anadromous fishes in the stream. The richness of nonnative species decreased at every site following the Accord, while the richness of native species increased or stayed constant. At the three most upstream sites, native species richness increased over time and ultimately exceeded nonnative richness. Native assemblage recovery was strongest upriver, closer to flow releases and habitat restoration activities, and decreased longitudinally downstream. Rank-abundance curves through time revealed that, while species evenness was low throughout the study, dominance shifted from nonnative to native species in the upstream sites coincident with rehabilitation efforts. Mean rank shifts decreased following flow rehabilitation; thus the assemblage became increasingly stable over time following flow rehabilitation. Putah Creek's rehabilitation may represent a model for others interested in improving endemic freshwater communities in degraded ecosystems.