No need for custom parser. Right now, you can use custom format with reviver and serializer functions (2nd arg to serialize / parse). Say, have them serialize as string `^bigint[\d]+$` and revive appropriately.
Given that `int + bigint` throws in JS, this exact thing will never happen imo, otherwise, you can't differentiate between types.
Maybe something like `{"x": 10n}`, but at that point, might as well do reviver, especially if you already agreed on a format and use some to parse out class instances vs plain objects.
e.g. so that {"foo": 1234567890123456789012345678901234567890} doesn't lose precision.