Fed-CausalDiff introduces a federated causal diffusion framework that enables do-simulation and policy evaluation in decentralized settings. It decomposes latent state evolution into global and local components, allowing decoupled synchronisation to reduce communication cost while maintaining accurate causal inference.