diff --git a/headroom/integrations/litellm_callback.py b/headroom/integrations/litellm_callback.py
index d5d54cacb..c88078069 100644
--- a/headroom/integrations/litellm_callback.py
+++ b/headroom/integrations/litellm_callback.py
@@ -94,6 +94,18 @@ class HeadroomCallback(_CustomLogger):
         """Whether cloud compression is enabled."""
         return self._api_key is not None
 
+    async def aclose(self) -> None:
+        """Close the shared cloud HTTP client, if it was initialized.
+
+        Applications using LiteLLM should await this method during their async
+        shutdown lifecycle. It is safe to call when cloud mode was not used or
+        after the client has already been closed.
+        """
+        client = self._client
+        self._client = None
+        if client is not None:
+            await client.aclose()
+
     async def async_pre_call_hook(
         self,
         user_api_key_dict: Any = None,
diff --git a/tests/test_integrations/test_litellm_callback.py b/tests/test_integrations/test_litellm_callback.py
index 50f698316..ddf71d4da 100644
--- a/tests/test_integrations/test_litellm_callback.py
+++ b/tests/test_integrations/test_litellm_callback.py
@@ -5,6 +5,7 @@ from __future__ import annotations
 import importlib
 import inspect
 from pathlib import Path
+from unittest.mock import AsyncMock, MagicMock
 
 import pytest
 
@@ -52,3 +53,30 @@ class TestHeadroomCallbackPostCallSuccessHook:
             response=sentinel,
         )
         assert result is sentinel
+
+
+class TestHeadroomCallbackClientLifecycle:
+    """Cloud client cleanup must be explicit and safe to repeat."""
+
+    @pytest.mark.asyncio
+    async def test_aclose_closes_and_clears_initialized_client(self) -> None:
+        cb = HeadroomCallback(api_key="hdr_test")
+        client = MagicMock()
+        client.aclose = AsyncMock()
+        cb._client = client
+
+        await cb.aclose()
+
+        client.aclose.assert_awaited_once_with()
+        assert cb._client is None
+
+        await cb.aclose()
+        client.aclose.assert_awaited_once_with()
+
+    @pytest.mark.asyncio
+    async def test_aclose_without_initialized_client_is_a_noop(self) -> None:
+        cb = HeadroomCallback(api_key="hdr_test")
+
+        await cb.aclose()
+
+        assert cb._client is None