Autots

AutoTS provides an MCP server interface for a Python time series forecasting package that enables rapid deployment of high-accuracy predictions at scale using dozens of forecasting models (naive, statistical, machine learning, and deep learning) and over 30 time series-specific transforms compatible with scikit-learn conventions. The server exposes tools for multivariate and probabilistic forecasting, automated model selection via genetic algorithms, horizontal and mosaic ensemble methods, cross-validation, and regressor generation—all operating directly on Pandas DataFrames without proprietary object conversion. AutoTS solves the problem of automating time series model selection and optimization for large datasets, having demonstrated competition-winning performance in the M6 forecasting competition.