Abstract: The invention concerns a method of providing a multi-modal dialog between a multi-modal application (5) and a user (6) communicating with the multi-modal application (5) via a client (4) suited to exchange and present documents (7) encoded in standard or extended hyper text mark-up language. The invention further concerns a proxy-server (2) for executing this method. The multi-modal dialog between the multi-modal application (5) and the user (6) is established through the proxy-server (2) interacting with the client (4) via exchange of information encoded in standard or extended hyper text mark-up language. The proxy-server (2) retrieves at least one additional resource of modality requested within the multi-modal dialog. The proxy-server (2) composes a multi-modal interaction with the user (6) based on standard or extended hyper text mark-up language interactions with the client (6) and on the retrieved additional resources (71, 72, 73, 31, 32, 33).