diff --git a/docs/combininglock.md b/docs/combininglock.md
index b951b7ae8..6aba2a41c 100644
--- a/docs/combininglock.md
+++ b/docs/combininglock.md
@@ -222,7 +222,7 @@ We can now provide our new version of the `with` function that uses the combinin
 template<typename F>
 inline void with(CombiningLock& lock, F&& f)
 {
-  CombiningLockNodeTempl<F> node(std::forward<F>(f));
+  CombiningLockNodeTempl<F> node(stl::forward<F>(f));
 
   // **************ACQUIRE**************
   // Add ourselves to the end of the queue.
diff --git a/src/snmalloc/aal/aal.h b/src/snmalloc/aal/aal.h
index 626cbf68b..3bbf63404 100644
--- a/src/snmalloc/aal/aal.h
+++ b/src/snmalloc/aal/aal.h
@@ -16,8 +16,9 @@
 #    define SNMALLOC_TICK_USE_CLOCK_GETTIME
 #  endif
 #endif
+#include "snmalloc/stl/utility.h"
+
 #include <stdint.h>
-#include <utility>
 
 #ifndef SNMALLOC_TICK_USE_CLOCK_GETTIME
 #  include <chrono>
diff --git a/src/snmalloc/backend/backend.h b/src/snmalloc/backend/backend.h
index ee170c38f..6a3fa8505 100644
--- a/src/snmalloc/backend/backend.h
+++ b/src/snmalloc/backend/backend.h
@@ -87,7 +87,7 @@ namespace snmalloc
      *   (remote, sizeclass, slab_metadata)
      * where slab_metadata, is the second element of the pair return.
      */
-    static std::pair<capptr::Chunk<void>, SlabMetadata*> alloc_chunk(
+    static stl::Pair<capptr::Chunk<void>, SlabMetadata*> alloc_chunk(
       LocalState& local_state,
       size_t size,
       uintptr_t ras,
diff --git a/src/snmalloc/backend/fixedglobalconfig.h b/src/snmalloc/backend/fixedglobalconfig.h
index 97318c5aa..5873ce2c5 100644
--- a/src/snmalloc/backend/fixedglobalconfig.h
+++ b/src/snmalloc/backend/fixedglobalconfig.h
@@ -1,8 +1,11 @@
 #pragma once
 
 #include "../backend_helpers/backend_helpers.h"
+#include "snmalloc/stl/type_traits.h"
 #include "standard_range.h"
 
+#include <stddef.h>
+
 namespace snmalloc
 {
   /**
diff --git a/src/snmalloc/backend_helpers/pagemap.h b/src/snmalloc/backend_helpers/pagemap.h
index 6a0cc5965..a6c909069 100644
--- a/src/snmalloc/backend_helpers/pagemap.h
+++ b/src/snmalloc/backend_helpers/pagemap.h
@@ -3,8 +3,7 @@
 #include "../ds/ds.h"
 #include "../mem/mem.h"
 #include "snmalloc/stl/atomic.h"
-
-#include <utility>
+#include "snmalloc/stl/utility.h"
 
 namespace snmalloc
 {
@@ -115,7 +114,7 @@ namespace snmalloc
      */
     template<bool fixed_range_ = fixed_range>
     static SNMALLOC_FAST_PATH
-      stl::enable_if_t<fixed_range_, std::pair<address_t, address_t>>
+      stl::enable_if_t<fixed_range_, stl::Pair<address_t, address_t>>
       get_bounds()
     {
       static_assert(fixed_range_ == fixed_range, "Don't set SFINAE parameter!");
diff --git a/src/snmalloc/ds/combininglock.h b/src/snmalloc/ds/combininglock.h
index e6e796dfd..190c90deb 100644
--- a/src/snmalloc/ds/combininglock.h
+++ b/src/snmalloc/ds/combininglock.h
@@ -256,7 +256,7 @@ namespace snmalloc
           reinterpret_cast<CombiningLockNodeTempl*>(self);
         self_templ->f();
       }),
-      f(std::forward<F>(f_))
+      f(stl::forward<F>(f_))
     {
       attach_slow(lock);
     }
@@ -290,6 +290,6 @@ namespace snmalloc
 
     // There is contention for the lock, we need to take the slow path
     // with the queue.
-    CombiningLockNodeTempl<F> node(lock, std::forward<F>(f));
+    CombiningLockNodeTempl<F> node(lock, stl::forward<F>(f));
   }
 } // namespace snmalloc
diff --git a/src/snmalloc/ds/pagemap.h b/src/snmalloc/ds/pagemap.h
index 1d285466f..bf1e8ffaf 100644
--- a/src/snmalloc/ds/pagemap.h
+++ b/src/snmalloc/ds/pagemap.h
@@ -115,7 +115,7 @@ namespace snmalloc
      * Returns usable range after pagemap has been allocated
      */
     template<bool has_bounds_ = has_bounds>
-    stl::enable_if_t<has_bounds_, std::pair<void*, size_t>>
+    stl::enable_if_t<has_bounds_, stl::Pair<void*, size_t>>
     init(void* b, size_t s)
     {
       SNMALLOC_ASSERT(!is_initialised());
@@ -250,7 +250,7 @@ namespace snmalloc
     }
 
     template<bool has_bounds_ = has_bounds>
-    stl::enable_if_t<has_bounds_, std::pair<address_t, size_t>> get_bounds()
+    stl::enable_if_t<has_bounds_, stl::Pair<address_t, size_t>> get_bounds()
     {
       SNMALLOC_ASSERT(is_initialised());
 
diff --git a/src/snmalloc/ds_aal/flaglock.h b/src/snmalloc/ds_aal/flaglock.h
index f98879904..e8dee95bd 100644
--- a/src/snmalloc/ds_aal/flaglock.h
+++ b/src/snmalloc/ds_aal/flaglock.h
@@ -24,7 +24,7 @@ namespace snmalloc
     constexpr DebugFlagWord() = default;
 
     template<typename... Args>
-    constexpr DebugFlagWord(Args&&... args) : flag(std::forward<Args>(args)...)
+    constexpr DebugFlagWord(Args&&... args) : flag(stl::forward<Args>(args)...)
     {}
 
     /**
@@ -87,7 +87,7 @@ namespace snmalloc
 
     template<typename... Args>
     constexpr ReleaseFlagWord(Args&&... args)
-    : flag(std::forward<Args>(args)...)
+    : flag(stl::forward<Args>(args)...)
     {}
 
     void set_owner() {}
diff --git a/src/snmalloc/ds_core/helpers.h b/src/snmalloc/ds_core/helpers.h
index 753b8edf9..2ff314b13 100644
--- a/src/snmalloc/ds_core/helpers.h
+++ b/src/snmalloc/ds_core/helpers.h
@@ -3,6 +3,7 @@
 #include "bits.h"
 #include "snmalloc/ds_core/defines.h"
 #include "snmalloc/stl/type_traits.h"
+#include "snmalloc/stl/utility.h"
 
 #include <array>
 #include <stddef.h>
@@ -357,8 +358,8 @@ namespace snmalloc
 
         if (fmt[0] == '{' && fmt[1] == '}')
         {
-          append(std::forward<Head>(head));
-          return append(fmt + 2, std::forward<Tail>(tail)...);
+          append(stl::forward<Head>(head));
+          return append(fmt + 2, stl::forward<Tail>(tail)...);
         }
 
         append_char(*fmt);
@@ -374,7 +375,7 @@ namespace snmalloc
     SNMALLOC_FAST_PATH MessageBuilder(const char* fmt, Args&&... args)
     {
       buffer[SafeLength] = 0;
-      append(fmt, std::forward<Args>(args)...);
+      append(fmt, stl::forward<Args>(args)...);
       append_char('\0');
     }
 
diff --git a/src/snmalloc/ds_core/seqset.h b/src/snmalloc/ds_core/seqset.h
index 65b188c31..6046bca70 100644
--- a/src/snmalloc/ds_core/seqset.h
+++ b/src/snmalloc/ds_core/seqset.h
@@ -3,6 +3,7 @@
 #include "../aal/aal.h"
 #include "../ds_core/ds_core.h"
 #include "snmalloc/stl/type_traits.h"
+#include "snmalloc/stl/utility.h"
 
 #include <stdint.h>
 
@@ -98,7 +99,7 @@ namespace snmalloc
     SNMALLOC_FAST_PATH bool is_empty()
     {
       static_assert(
-        stl::is_same_v<Node, decltype(std::declval<T>().node)>,
+        stl::is_same_v<Node, decltype(stl::declval<T>().node)>,
         "T->node must be Node for T");
       head.invariant();
       return head.next == &head;
diff --git a/src/snmalloc/mem/backend_concept.h b/src/snmalloc/mem/backend_concept.h
index ee288d37d..12fd0c017 100644
--- a/src/snmalloc/mem/backend_concept.h
+++ b/src/snmalloc/mem/backend_concept.h
@@ -4,7 +4,7 @@
 #  include "../ds/ds.h"
 #  include "sizeclasstable.h"
 
-#  include <cstddef>
+#  include <stddef.h>
 
 namespace snmalloc
 {
@@ -107,7 +107,7 @@ namespace snmalloc
       {
         Backend::alloc_chunk(local_state, size, ras, sizeclass)
         } -> ConceptSame<
-          std::pair<capptr::Chunk<void>, typename Backend::SlabMetadata*>>;
+          stl::Pair<capptr::Chunk<void>, typename Backend::SlabMetadata*>>;
     } &&
     requires(LocalState* local_state, size_t size) {
       {
diff --git a/src/snmalloc/mem/backend_wrappers.h b/src/snmalloc/mem/backend_wrappers.h
index 9e1d3370f..4eb39465f 100644
--- a/src/snmalloc/mem/backend_wrappers.h
+++ b/src/snmalloc/mem/backend_wrappers.h
@@ -46,8 +46,8 @@ namespace snmalloc
     constexpr SNMALLOC_FAST_PATH auto has_domesticate(int) -> stl::enable_if_t<
       stl::is_same_v<
         decltype(Config::capptr_domesticate(
-          std::declval<typename Config::LocalState*>(),
-          std::declval<CapPtr<T, B>>())),
+          stl::declval<typename Config::LocalState*>(),
+          stl::declval<CapPtr<T, B>>())),
         CapPtr<
           T,
           typename B::template with_wildness<
diff --git a/src/snmalloc/mem/freelist.h b/src/snmalloc/mem/freelist.h
index 75c96f27a..7e5c9ca65 100644
--- a/src/snmalloc/mem/freelist.h
+++ b/src/snmalloc/mem/freelist.h
@@ -906,7 +906,7 @@ namespace snmalloc
       template<bool RANDOM_ = RANDOM>
       stl::enable_if_t<
         !RANDOM_,
-        std::pair<
+        stl::Pair<
           Object::BHeadPtr<BView, BQueue>,
           Object::BHeadPtr<BView, BQueue>>>
       extract_segment(const FreeListKey& key, address_t key_tweak)
diff --git a/src/snmalloc/mem/localalloc.h b/src/snmalloc/mem/localalloc.h
index 2cdefc81c..d30d4149d 100644
--- a/src/snmalloc/mem/localalloc.h
+++ b/src/snmalloc/mem/localalloc.h
@@ -21,8 +21,9 @@
 #  include "external_alloc.h"
 #endif
 
+#include "snmalloc/stl/utility.h"
+
 #include <string.h>
-#include <utility>
 
 namespace snmalloc
 {
diff --git a/src/snmalloc/mem/metadata.h b/src/snmalloc/mem/metadata.h
index 9821dc15d..2cbe630bd 100644
--- a/src/snmalloc/mem/metadata.h
+++ b/src/snmalloc/mem/metadata.h
@@ -620,7 +620,7 @@ namespace snmalloc
      * available objects for this slab metadata.
      */
     template<typename Domesticator>
-    static SNMALLOC_FAST_PATH std::pair<freelist::HeadPtr, bool>
+    static SNMALLOC_FAST_PATH stl::Pair<freelist::HeadPtr, bool>
     alloc_free_list(
       Domesticator domesticate,
       FrontendSlabMetadata* meta,
diff --git a/src/snmalloc/mem/remoteallocator.h b/src/snmalloc/mem/remoteallocator.h
index 193d753b6..3e021297d 100644
--- a/src/snmalloc/mem/remoteallocator.h
+++ b/src/snmalloc/mem/remoteallocator.h
@@ -104,7 +104,7 @@ namespace snmalloc
     }
 
     template<SNMALLOC_CONCEPT(IsConfigLazy) Config, typename Domesticator_queue>
-    SNMALLOC_FAST_PATH static std::pair<freelist::HeadPtr, uint16_t>
+    SNMALLOC_FAST_PATH static stl::Pair<freelist::HeadPtr, uint16_t>
     open_free_ring(
       capptr::Alloc<BatchedRemoteMessage> m,
       size_t objsize,
@@ -242,7 +242,7 @@ namespace snmalloc
     }
 
     template<SNMALLOC_CONCEPT(IsConfigLazy) Config, typename Domesticator_queue>
-    SNMALLOC_FAST_PATH static std::pair<freelist::HeadPtr, uint16_t>
+    SNMALLOC_FAST_PATH static stl::Pair<freelist::HeadPtr, uint16_t>
     open_free_ring(
       capptr::Alloc<SingletonRemoteMessage> m,
       size_t,
diff --git a/src/snmalloc/pal/pal.h b/src/snmalloc/pal/pal.h
index a194e7934..6fbb8f347 100644
--- a/src/snmalloc/pal/pal.h
+++ b/src/snmalloc/pal/pal.h
@@ -168,14 +168,14 @@ namespace snmalloc
   template<size_t BufferSize, typename... Args>
   [[noreturn]] inline void report_fatal_error(Args... args)
   {
-    MessageBuilder<BufferSize> msg{std::forward<Args>(args)...};
+    MessageBuilder<BufferSize> msg{stl::forward<Args>(args)...};
     DefaultPal::error(msg.get_message());
   }
 
   template<size_t BufferSize, typename... Args>
   inline void message(Args... args)
   {
-    MessageBuilder<BufferSize> msg{std::forward<Args>(args)...};
+    MessageBuilder<BufferSize> msg{stl::forward<Args>(args)...};
     MessageBuilder<BufferSize> msg_tid{
       "{}: {}", debug_get_tid(), msg.get_message()};
     DefaultPal::message(msg_tid.get_message());
diff --git a/src/snmalloc/pal/pal_posix.h b/src/snmalloc/pal/pal_posix.h
index 0d50f36a0..1b1b3b8ab 100644
--- a/src/snmalloc/pal/pal_posix.h
+++ b/src/snmalloc/pal/pal_posix.h
@@ -5,6 +5,8 @@
 #if defined(SNMALLOC_BACKTRACE_HEADER)
 #  include SNMALLOC_BACKTRACE_HEADER
 #endif
+#include "snmalloc/stl/utility.h"
+
 #include <errno.h>
 #include <fcntl.h>
 #include <limits.h>
@@ -15,7 +17,6 @@
 #include <sys/mman.h>
 #include <sys/uio.h>
 #include <unistd.h>
-#include <utility>
 
 #if __has_include(<sys/random.h>)
 #  include <sys/random.h>
diff --git a/src/snmalloc/stl/cxx/utility.h b/src/snmalloc/stl/cxx/utility.h
new file mode 100644
index 000000000..0853239b2
--- /dev/null
+++ b/src/snmalloc/stl/cxx/utility.h
@@ -0,0 +1,13 @@
+#include <utility>
+
+namespace snmalloc
+{
+  namespace stl
+  {
+    using std::declval;
+    using std::forward;
+    using std::move;
+    template<class T1, class T2>
+    using Pair = std::pair<T1, T2>;
+  } // namespace stl
+} // namespace snmalloc
diff --git a/src/snmalloc/stl/gnu/type_traits.h b/src/snmalloc/stl/gnu/type_traits.h
index 7b2d7691a..5ae96685f 100644
--- a/src/snmalloc/stl/gnu/type_traits.h
+++ b/src/snmalloc/stl/gnu/type_traits.h
@@ -390,5 +390,21 @@ namespace snmalloc
     template<class T>
     using add_const_t = const T;
 
+#if __has_builtin(__is_lvalue_reference)
+    template<class T>
+    inline constexpr bool is_lvalue_reference_v = __is_lvalue_reference(T);
+#else
+    template<typename T>
+    struct is_lvalue_reference : public false_type
+    {};
+
+    template<typename T>
+    struct is_lvalue_reference<T&> : public true_type
+    {};
+
+    template<typename T>
+    inline constexpr bool is_lvalue_reference_v = is_lvalue_reference<T>::value;
+#endif
+
   } // namespace stl
 } // namespace snmalloc
diff --git a/src/snmalloc/stl/gnu/utility.h b/src/snmalloc/stl/gnu/utility.h
new file mode 100644
index 000000000..193613097
--- /dev/null
+++ b/src/snmalloc/stl/gnu/utility.h
@@ -0,0 +1,68 @@
+#pragma once
+
+#include "snmalloc/stl/type_traits.h"
+
+// This is used by clang to provide better analysis of lifetimes.
+#if __has_cpp_attribute(_Clang::__lifetimebound__)
+#  define SNMALLOC_LIFETIMEBOUND [[_Clang::__lifetimebound__]]
+#else
+#  define SNMALLOC_LIFETIMEBOUND
+#endif
+
+namespace snmalloc
+{
+  namespace stl
+  {
+    template<class T>
+    [[nodiscard]] inline constexpr T&&
+    forward(remove_reference_t<T>& ref) noexcept
+    {
+      return static_cast<T&&>(ref);
+    }
+
+    template<class T>
+    [[nodiscard]] inline constexpr T&&
+    forward(SNMALLOC_LIFETIMEBOUND remove_reference_t<T>&& ref) noexcept
+    {
+      static_assert(
+        !is_lvalue_reference_v<T>, "cannot forward an rvalue as an lvalue");
+      return static_cast<T&&>(ref);
+    }
+
+    template<class T>
+    [[nodiscard]] inline constexpr remove_reference_t<T>&&
+    move(SNMALLOC_LIFETIMEBOUND T&& ref) noexcept
+    {
+#ifdef __clang__
+      using U [[gnu::nodebug]] = remove_reference_t<T>;
+#else
+      using U = remove_reference_t<T>;
+#endif
+      return static_cast<U&&>(ref);
+    }
+
+#pragma GCC diagnostic push
+#pragma GCC diagnostic ignored "-Wdeprecated"
+    // First try instantiation with reference types, then fall back to value
+    // types. Use int to prioritize reference types.
+    template<class T>
+    T&& declval_impl(int);
+    template<class T>
+    T declval_impl(long);
+#pragma GCC diagnostic pop
+
+    template<class T>
+    constexpr inline decltype(declval_impl<T>(0)) declval() noexcept
+    {
+      static_assert(
+        !is_same_v<T, T>, "declval cannot be used in an evaluation context");
+    }
+
+    template<class T1, class T2>
+    struct Pair
+    {
+      T1 first;
+      T2 second;
+    };
+  } // namespace stl
+} // namespace snmalloc
diff --git a/src/snmalloc/stl/utility.h b/src/snmalloc/stl/utility.h
new file mode 100644
index 000000000..78e3e6af9
--- /dev/null
+++ b/src/snmalloc/stl/utility.h
@@ -0,0 +1,9 @@
+#pragma once
+
+#include "snmalloc/stl/common.h"
+
+#if SNMALLOC_USE_SELF_VENDORED_STL
+#  include "snmalloc/stl/gnu/utility.h"
+#else
+#  include "snmalloc/stl/cxx/utility.h"
+#endif