Reservoir computing (RC) is a computationally efficient framework for time-series learning in which the internal recurrent connections (the reservoir) are fixed and only a linear readout is trained. By nonlinearly projecting input time series into high-dimensional reservoir states and casting training as regularized linear regression, RC is well suited to ecological dynamics that are nonlinear, multivariate, and often subject to noise, missing data, and external disturbances. In this talk, I will provide an intuitive overview of (1) the basic mechanism of RC, (2) its connections to nonlinear time-series analysis, and (3) key applications in ecology, ranging from short-term forecasting to inferring interaction structure (with links to causal discovery), estimating dynamical indicators such as the maximal Lyapunov exponent, and evaluating the computational capacity of ecosystems. I will also briefly highlight practical implementation tips and caveats, and discuss future prospects for RC in ecology.