CADRE

Connected Models (aka CADRE: Composable Atomic Deterministic Registry Engine) is a repository of mathematical equations and models from a variety of domains, that can be assembled and executed in real time. These models were curated by specially trained Muse agents, which extracted equations from textbooks, papers, and websites. They were then audited by specially trained auditor Muse agents. This mining process is ongoing and expected to continue, in additional domains. Your LLM should reason, not perform math. One model's output supplies another's input, replacing arbitrary assumptions with explicit traceable dependencies. CADRE parses your query into typed facts with exact quotes, backward-chains over a registry of eligible versioned models, executes deterministically with ratio-unit conversion, guard checks, and exact meaning/context matching, then saves an immutable run. Connected-models is available via web interface, Python CLI, agent SKILL, and MCP server.