We identify the causal effect in a multi-domain setting with unobserved confounders using proxies, given only measurements of the proxy in the target domain, and provide consistent estimators.