An MCP server that evaluates whether retrieval methods and AI outputs are grounded in long narrative manuscripts by retrieving evidence and scoring coverage deterministically, without external model A…