diff --git a/include/ftl/fake_guard.h b/include/ftl/fake_guard.h
new file mode 100644
index 0000000..bacd1b2
--- /dev/null
+++ b/include/ftl/fake_guard.h
@@ -0,0 +1,90 @@
+/*
+ * Copyright 2022 The Android Open Source Project
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#pragma once
+
+#define FTL_ATTRIBUTE(a) __attribute__((a))
+
+namespace android::ftl {
+
+// Granular alternative to [[clang::no_thread_safety_analysis]]. Given a std::mutex-like object,
+// FakeGuard suppresses enforcement of thread-safe access to guarded variables within its scope.
+// While FakeGuard is scoped to a block, there are macro shorthands for a single expression, as
+// well as function/lambda scope (though calls must be indirect, e.g. virtual or std::function):
+//
+//   struct {
+//     std::mutex mutex;
+//     int x FTL_ATTRIBUTE(guarded_by(mutex)) = -1;
+//
+//     int f() {
+//       {
+//         ftl::FakeGuard guard(mutex);
+//         x = 0;
+//       }
+//
+//       return FTL_FAKE_GUARD(mutex, x + 1);
+//     }
+//
+//      std::function<int()> g() const {
+//        return [this]() FTL_FAKE_GUARD(mutex) { return x; };
+//      }
+//   } s;
+//
+//   assert(s.f() == 1);
+//   assert(s.g()() == 0);
+//
+// An example of a situation where FakeGuard helps is a mutex that guards writes on Thread 1, and
+// reads on Thread 2. Reads on Thread 1, which is the only writer, need not be under lock, so can
+// use FakeGuard to appease the thread safety analyzer. Another example is enforcing and documenting
+// exclusive access by a single thread. This is done by defining a global constant that represents a
+// thread context, and annotating guarded variables as if it were a mutex (though without any effect
+// at run time):
+//
+//   constexpr class [[clang::capability("mutex")]] {
+//   } kMainThreadContext;
+//
+template <typename Mutex>
+struct [[clang::scoped_lockable]] FakeGuard final {
+  explicit FakeGuard(const Mutex& mutex) FTL_ATTRIBUTE(acquire_capability(mutex)) {}
+  [[clang::release_capability()]] ~FakeGuard() {}
+
+  FakeGuard(const FakeGuard&) = delete;
+  FakeGuard& operator=(const FakeGuard&) = delete;
+};
+
+}  // namespace android::ftl
+
+// TODO: Enable in C++23 once standard attributes can be used on lambdas.
+#if 0
+#define FTL_FAKE_GUARD1(mutex) [[using clang: acquire_capability(mutex), release_capability(mutex)]]
+#else
+#define FTL_FAKE_GUARD1(mutex)             \
+  FTL_ATTRIBUTE(acquire_capability(mutex)) \
+  FTL_ATTRIBUTE(release_capability(mutex))
+#endif
+
+// The parentheses around `expr` are needed to deduce an lvalue or rvalue reference.
+#define FTL_FAKE_GUARD2(mutex, expr)            \
+  [&]() -> decltype(auto) {                     \
+    const android::ftl::FakeGuard guard(mutex); \
+    return (expr);                              \
+  }()
+
+#define FTL_MAKE_FAKE_GUARD(arg1, arg2, guard, ...) guard
+
+// The void argument suppresses a warning about zero variadic macro arguments.
+#define FTL_FAKE_GUARD(...) \
+  FTL_MAKE_FAKE_GUARD(__VA_ARGS__, FTL_FAKE_GUARD2, FTL_FAKE_GUARD1, void)(__VA_ARGS__)