ABSTRACTNeotropical freshwater crabs are understudied but ecologically important consumers, with largely unknown responses to land use change and watershed urbanization. We used capture–mark–recapture methods to quantify Ptychophallus tumimanus (Pseudothelphusidae) freshwater crab populations across 20 headwater stream sites spanning a range of land uses, including forest reserves and an urban center in Monteverde, Costa Rica. We estimated site‐specific crab abundance and density using hierarchical Bayesian Markov chain Monte Carlo models and estimated the effects of potential covariates (elevation, water temperature, conductivity, % cobble, mean substrate size, channel width, depth, stream velocity, pH, % canopy cover) on crab abundances across sites. Crab abundances ranged from ~2 (95% CI: 0–9) to 143 (95% CI: 98–207) crabs across stream reaches. Densities ranged from 0.06 (95% CI: 0–0.25) to 5.21 (95% CI: 3.38–8.00) crabs/m of stream length. Probability of capture was estimated as 9.7% for a 30 min search time. Streams in undisturbed forested watersheds had higher crab abundances than sites near human development, where we found few to no crabs. Of the 10 predictor variables, conductivity and % cobble substrate best explained differences in P. tumimanus abundances across sites. Urban streams had low crab abundances and higher conductivities than forested streams, which suggests that this crab species may be sensitive to in‐stream effects of urbanization. This study demonstrates that capture–mark–recapture methods can be useful for estimating and understanding how crab abundances may vary across streams.