Clio builds memory by mapping how concepts connect — not just that they do. Each link is typed: drives, part_of, enables, measures, analogy. When a query arrives, retrieval follows causal chains through the structure rather than guessing from similar text. The result is a local knowledge graph extracted once from your files, reused across every future query, at a fraction of the token cost - and optimised for problem solving by reasoning through the structure of the issue, not around it.