mcp -scikit-learn

This MCP server implements a Scikit-learn workflow exposed through the MCP interface. It provides capabilities for training and evaluating Scikit-learn models, handling datasets with preprocessing steps, persisting and loading models, performing feature engineering and selection, computing evaluation metrics, and supporting cross-validation and hyperparameter tuning. You can interact with the server through the MCP inspector or any MCP client by sending context-aware requests that trigger training runs, make predictions, evaluate models, and manage datasets. The server is designed to integrate into a broader MCP workflow so you can chain data preparation, model training, evaluation, and deployment tasks in a repeatable, context-aware manner.

Prerequisites:

• Python 3.8+
• Virtual environment support (optional but recommended)
• Git

Step-by-step:

1. Clone the repository: git clone https://github.com/yourusername/mcp-server-scikit-learn.git cd mcp-server-scikit-learn

2. Create and activate a virtual environment (recommended): python -m venv .venv source .venv/bin/activate # macOS/Linux .venv\Scripts\activate # Windows

3. Install dependencies (editable with dev dependencies): pip install -e ".[dev]"

4. Run tests (optional to verify setup): pytest -s -v tests/

5. Start the MCP server via the MCP inspector or as a standalone service as described in the README. For example, integrate with MCP inspector using the provided uv-based configuration.

Note: If you plan to run in a virtual environment or container, ensure your PATH and environment variables point to the correct Python and dependencies.

Tips and caveats:

• The README indicates this project is not yet configured for ephemeral environments (uvx). Plan to run locally first to validate the workflow.
• When integrating as an MCP server, replace the placeholder /path/to/mcp-server-scikit-learn with the actual directory containing the server code.
• Ensure Python dependencies are installed in the active environment before launching the server.
• For dataset handling and preprocessing, consider configuring data pipelines within your MCP requests to streamline training and evaluation.
• If you enable hyperparameter tuning or cross-validation, be mindful of compute time and resource usage in your environment.
• Environment variables can be used to control logging verbosity or to point to model storage locations; add them to the mcp_config.env object as needed.