Decentralized execution of multi-agent LLM workflows requires deterministic memory persistence and local fallback execution without relying on unstable external cloud APIs. Needed unified IPC pipeline and schema validation across heterogeneous model agents.