AutoTS

AutoTS is an automated time series forecasting library that selects the best model for your data by running multiple algorithms and comparing their performance. The MCP server wraps this capability so your AI agent can trigger forecasting jobs, explore results, and iterate on predictions conversationally. Instead of writing Python scripts to test different forecasting approaches, you describe what you need and the agent handles the pipeline. The library behind AutoTS supports over 30 forecasting models including statistical methods (ARIMA, ETS), machine learning approaches (XGBoost, LightGBM), and deep learning models. It automatically handles data preprocessing, model selection, hyperparameter tuning, and ensemble generation. Through the MCP interface, your agent can load datasets, configure forecast horizons, and retrieve predictions with confidence intervals. With 1,300+ stars and active development, AutoTS has a solid reputation in the Python data science community. The MCP layer adds a conversational interface on top, which is particularly useful for non-technical stakeholders who want to explore forecasts without writing code. The trade-off is that automated model selection can be slow for large datasets, and you lose fine-grained control over model configuration.