Retrieval Relevance Scorer MCP Connector for Claude

This MCP server provides a deterministic scoring engine to solve the noise and hallucination problems in RAG (Retrieval-Augmented Generation). It evaluates the relationship between a query and retrieved documents using three mathematical metrics: Jaccard Similarity for keyword overlap, TF-IDF Cosine Similarity for term importance, and Query Term Coverage for information density. Use score_documents to filter out irrelevant context, get_scoring_config to inspect weight distributions, or analyze_coverage_gap to diagnose why specific terms are missing from your retrieval set. It is designed to ensure only high-signal text reaches your AI client.