<xstd/ints/cstdint.hpp>

#define XSTD_HAS_INT128 see below
#define XSTD_HAS_BIT_INT see below

namespace xstd {
  // where XSTD_HAS_INT128 is defined
  using int128 = see below;
  using uint128 = see below;
  template<> struct make_unsigned<int128>;
  template<> struct make_signed<uint128>;

  // where XSTD_HAS_BIT_INT is defined
  inline constexpr size_t bit_int_max_width = __BITINT_MAXWIDTH__;
  template<size_t N>
    requires (2 <= N and N <= bit_int_max_width)
    using bit_int = signed _BitInt(N);
  template<size_t N>
    requires (2 <= N and N <= bit_int_max_width)
    using bit_uint = unsigned _BitInt(N);
}

Header	Declares
<xstd/ints/cstdint/bit_int.hpp>	bit_int, bit_uint
<xstd/ints/cstdint/int128.hpp>	int128, uint128

The umbrella exports <xstd/ints/cstdint/int128.hpp> and <xstd/ints/cstdint/bit_int.hpp>. Every type it names is half of a pair: each transformation leads to the other half and back, and the unsigned half has one more digit than the signed half, cv-qualified or not.

<xstd/ints/cstdint/bit_int.hpp> exports <xstd/ints/limits/numeric_limits.hpp>, <xstd/ints/type_traits/make_signed.hpp> and <xstd/ints/type_traits/make_unsigned.hpp>, which declare the _BitInt specializations of numeric_limits, make_signed and make_unsigned, so including it is enough to ask them of the aliases.