The ‘mentalese’ the Chomskyans are looking for, is likely thought itself.
That’s not language properly speaking, and applies to lower animals, as well. Once we have language, we proceed to think in it, but that doesn’t mean that we couldn’t think without it. The idea that, since all languages are so similar, and since all children learn them so easily, then there must be an underlying ‘mental language’, makes a few non-provable conclusions based on a few non-provable assumptions, though it may fall short of outright begging the question. For one thing, though I love kids, their linguistic prowess is not impressive to me. Think what you might do if you had one-on-one instruction every day for four or five years with literally nothing else to occupy your mind and everything to gain for your efforts. Secondly, since when are all languages so similar? They may indeed all be coming closer together whether because of international English or the simple logic and proven effectiveness of S-V-O word order, but that is recent and tentative. There is a much longer history of languages categorized as synthetic/analytical, inflected, or agglutinative. There may be an even earlier period when languages were more similar. Nevertheless, if languages are indeed similar, there may be an even better reason for the phenomenon. They may all derive ultimately from the same parent language before they literally went separate ways.