"author:pranav797" matching MCP servers:

Enables Claude clients to score borrower default risk, explain predictions with SHAP, compare applicants, and inspect model metadata through natural conversation. Serves a trained XGBoost model locally over stdio or remotely over HTTP with API-key authentication, keeping confidence transparent by reporting which inputs fell back to training-set defaults.