The topos version of forcing with a poset $P$ regards $P$ as a category, forms the topos of presheaves on it, and then passes to the subtopos of double-negation sheaves. The presheaf topos amounts to strong forcing (which obeys only intuitionistic logic) and the sheaf subtopos amounts to the much more widely used weak forcing. This was, at least for the Cohen models violating CH, in the early work of Lawvere and Tierney, presented in Lawvere's paper "Quantifiers and sheaves" (Proceedings of the 1970 ICM in Nice. Big pdf (see p329), individual article). If I remember correctly, Marta Bunge worked out the topos version of the forcing that adjoins a Suslin tree (JPAA 1974, doi:10.1016/0022-4049(74)90020-6, see also the erratum).
More information about these topics is in work of John Bell (connecting it with Boolean-valued models, if I remember correctly) and in Peter Freyd's paper "All topoi are localic" (JPAA 1987, doi:10.1016/0022-4049(87)90042-9). Andre Scedrov and I included much of this material in the early, background sections of "Freyd's models for the axiom of choice" (Memoirs A.M.S. 404 (1989) --- this was before "404" became a widely known synonym for "file not found").