Ckg Mcp

Exposes structured knowledge graphs as MCP tools instead of embedding chunks into a vector store. You get four operations: list_domains() to browse 53 prebuilt graphs (langchain-core, glp1-obesity, calculus, etc), query_ckg() to extract subgraphs by depth, get_prerequisites() to walk dependency chains, and search_concepts() for lookup. Built for structural queries where RAG burns tokens inferring relationships that already exist in the graph. The benchmark claims 65x token efficiency on multi-hop prerequisite questions. Reach for this when you need blast radius analysis on codebases, prerequisite chains in clinical domains, or any scenario where concept dependencies matter more than semantic similarity.