We develop methods for robust Bayesian inference in structural vector autoregressions (SVARs) where the parameters of interest are set-identified using external instruments, or ‘proxy SVARs’. Set-identification in these models typically occurs when there are multiple instruments for multiple structural shocks. Existing Bayesian approaches to inference in proxy SVARs require researchers to specify a single prior over the model’s parameters, but, under set-identification, a component of the prior is never revised. We extend the robust Bayesian approach to inference in set-identified models proposed by Giacomini and Kitagawa (2018) – which allows researchers to relax potentially con-troversial point-identifying restrictions without having to specify an unrevisable prior – to proxy SVARs. We provide new results on the frequentist validity of the approach in proxy SVARs. We also explore the effect of instrument strength on inference about the identified set. We illustrate our approach by revisiting Mertens and Ravn (2013) and relaxing the assumption that they impose to obtain point identification.