# Skforecast Studio > Skforecast Studio (https://studio.skforecast.org) is a free, no-code web app for time series forecasting with machine learning. It runs the open-source Python library skforecast on a server and turns every step of the pipeline into plain Python code that the user can download as a script or a Jupyter notebook. Use it to go from a CSV or Excel file to a validated forecast without writing code, or to learn skforecast by seeing the code that each click generates. ## What it does - Load data: three demo datasets (monthly water demand with exogenous variables, monthly fuel consumption in Spain, hourly electricity demand in Victoria, Australia) or your own CSV or Excel file with a date column, a target column and optional exogenous columns. The frequency is inferred automatically. - Explore: summary statistics, interactive charts, seasonal decomposition, autocorrelation and partial autocorrelation plots. - Feature engineering: lags, rolling window statistics, calendar features and transformations. - Forecaster setup: recursive or direct multi-step forecasters with LightGBM, XGBoost, random forest, Ridge or linear regression, and statistical models (ARIMA, ETS). - Fit and predict: train, forecast ahead and get prediction intervals (bootstrapping or conformal). - Backtesting: walk-forward validation with refit, no refit, fixed train size, gap and fold stride, and error metrics per fold (MAE, MSE, MAPE, MASE and more). - Hyperparameter tuning and lag selection with Bayesian search (Optuna). - Explainability: feature importance and SHAP values. - Export: the Python code of the current step or the full pipeline, as a script or a Jupyter notebook, and the fitted forecaster. ## Limits and privacy - Free to use, no sign-up and no account. - Uploaded data and forecasts are never stored: the server is stateless and only keeps the data for the duration of each request. - Uploaded files up to 5 MB and 50,000 rows per dataset. Heavy operations (fit, backtesting, search, explainability) time out after 60 seconds. - Univariate forecasting of one series at a time, with optional exogenous variables. - For larger data or production use, run the exported Python code with skforecast on your own machine. ## Frequently asked questions ### Is Skforecast Studio free? Yes. Studio is free to use, with no sign-up and no account. It is built and maintained by the creators of skforecast, the open-source library it runs on. ### Do you store my data? No. Your dataset stays in your browser tab and is only held in the server memory while a request runs. It is never written to a database or a file, and it is never shared. Analytics cookies are only set if you accept them and never include your data. ### What are the limits? Files up to 5 MB and 50,000 rows, one series at a time with optional exogenous variables. Fit, backtesting, hyperparameter search and explainability stop after 60 seconds. For larger data, run the exported Python code on your own machine. ### How is it different from writing the code myself? It is the same code. Every step runs skforecast in Python and Studio writes the equivalent script as you go. Download it as a Python script or a Jupyter notebook and keep working without the app: Studio is the quick way to try ideas, the code is what you take to production. ### Can I use it at work? Yes. The exported code only uses open-source libraries with permissive licenses, such as skforecast, scikit-learn and pandas, so you can run it anywhere, including commercial projects. Studio itself is a free shared service without availability guarantees. If you need more privacy, higher limits or heavier workloads, contact us: we can deploy a dedicated instance for your team. ## Links that open a demo A link can open Studio with a demo dataset already loaded, on a given step: - https://studio.skforecast.org/?dataset=h2o : monthly water demand with exogenous variables, on the Explore step. - https://studio.skforecast.org/?dataset=fuel : monthly fuel consumption in Spain, on the Explore step. - https://studio.skforecast.org/?dataset=vic_electricity&step=fit : hourly electricity demand in Victoria, ready to fit a forecaster. `dataset` is one of `h2o`, `fuel`, `vic_electricity`. `step` is optional, one of `load`, `explore`, `features`, `forecaster`, `fit`, `backtesting`, `hyperparameters`, `explainability`, `export` (default `explore`). ## Documentation - [skforecast](https://skforecast.org): documentation of the Python library behind Studio. - [Introduction to forecasting](https://skforecast.org/latest/introduction-forecasting/introduction-forecasting): concepts used in Studio. - [Recursive multi-step forecasting](https://skforecast.org/latest/user_guides/autoregressive-forecaster) - [Direct multi-step forecasting](https://skforecast.org/latest/user_guides/direct-multi-step-forecasting) - [Statistical models](https://skforecast.org/latest/user_guides/forecasting-statistical-models) - [Backtesting](https://skforecast.org/latest/user_guides/backtesting.html) - [Hyperparameter tuning and lag selection](https://skforecast.org/latest/user_guides/hyperparameter-tuning-and-lags-selection.html) - [Explainability](https://skforecast.org/latest/user_guides/explainability) ## Optional - [skforecast on GitHub](https://github.com/skforecast/skforecast): source code of the library (BSD 3-Clause). - [skforecast on PyPI](https://pypi.org/project/skforecast/) - [Discord community](https://discord.gg/3V52qpNkuj)