diff --git a/Makefile b/Makefile
index 2dce900..4cb1bac 100644
--- a/Makefile
+++ b/Makefile
@@ -6,6 +6,12 @@ examples:
 	clang++ -flto -O3 -s *.o -o HelloWorld
 	rm *.o
 
+# HelloWorldThreaded (non-SSL, non-Zlib compile)
+	clang -DLIBUS_NO_SSL -flto -O3 -c -IuSockets/src uSockets/src/*.c uSockets/src/eventing/*.c
+	clang++ -DLIBUS_NO_SSL -DUWS_NO_ZLIB -flto -O3 -c -std=c++17 -Isrc -IuSockets/src examples/HelloWorldThreaded.cpp
+	clang++ -lpthread -flto -O3 -s *.o -o HelloWorldThreaded
+	rm *.o
+
 # EchoServer (non-SSL, non-Zlib compile)
 	clang -DLIBUS_NO_SSL -flto -O3 -c -IuSockets/src uSockets/src/*.c uSockets/src/eventing/*.c
 	clang++ -DLIBUS_NO_SSL -DUWS_NO_ZLIB -flto -O3 -c -std=c++17 -Isrc -IuSockets/src examples/EchoServer.cpp
diff --git a/examples/HelloWorldThreaded.cpp b/examples/HelloWorldThreaded.cpp
new file mode 100644
index 0000000..a35e9b5
--- /dev/null
+++ b/examples/HelloWorldThreaded.cpp
@@ -0,0 +1,28 @@
+#include "App.h"
+#include <thread>
+#include <algorithm>
+
+int main() {
+    /* Overly simple hello world app, using multiple threads */
+    std::vector<std::thread *> threads(std::thread::hardware_concurrency());
+
+    std::transform(threads.begin(), threads.end(), threads.begin(), [](std::thread *t) {
+        return new std::thread([]() {
+
+            uWS::App().get("/*", [](auto *res, auto *req) {
+                res->end("Hello world!");
+            }).listen(3000, [](auto *token) {
+                if (token) {
+                    std::cout << "Thread " << std::this_thread::get_id() << " listening on port " << 3000 << std::endl;
+                } else {
+                    std::cout << "Thread " << std::this_thread::get_id() << " failed to listen on port 3000" << std::endl;
+                }
+            }).run();
+
+        });
+    });
+
+    std::for_each(threads.begin(), threads.end(), [](std::thread *t) {
+        t->join();
+    });
+}
diff --git a/src/Loop.h b/src/Loop.h
index f030755..810a8a9 100644
--- a/src/Loop.h
+++ b/src/Loop.h
@@ -24,7 +24,7 @@
 #include <libusockets_new.h>
 
 #include <iostream>
-#include <atomic>
+#include <thread>
 
 namespace uWS {
 struct Loop {
@@ -86,9 +86,13 @@ private:
 public:
     /* Returns the default loop if called from one thread, or a dedicated per-thread loop if called from multiple threads */
     static Loop *defaultLoop() {
+        /* Lock this whole function */
+        static std::mutex m;
+        std::lock_guard<std::mutex> lock(m);
+
         /* Deliver and attach the default loop to the first thread who calls us */
         static thread_local bool ownsDefaultLoop;
-        static std::atomic<Loop *> defaultLoop;
+        static Loop *defaultLoop;
         if (!defaultLoop) {
             ownsDefaultLoop = true;
             defaultLoop = create(true);