diff --git a/litellm/integrations/custom_guardrail.py b/litellm/integrations/custom_guardrail.py
index 501185b207..1ca45f907e 100644
--- a/litellm/integrations/custom_guardrail.py
+++ b/litellm/integrations/custom_guardrail.py
@@ -119,11 +119,8 @@ class CustomGuardrail(CustomLogger):
         """
         if "guardrails" in data:
             return data["guardrails"]
-        metadata = data.get("metadata") or {}
-        requested_guardrails = metadata.get("guardrails") or []
-        if requested_guardrails:
-            return requested_guardrails
-        return requested_guardrails
+        metadata = data.get("litellm_metadata") or data.get("metadata", {})
+        return metadata.get("guardrails") or []
 
     def _guardrail_is_in_requested_guardrails(
         self,
diff --git a/litellm/proxy/anthropic_endpoints/endpoints.py b/litellm/proxy/anthropic_endpoints/endpoints.py
index a10a39a6a5..2de5ec1ee1 100644
--- a/litellm/proxy/anthropic_endpoints/endpoints.py
+++ b/litellm/proxy/anthropic_endpoints/endpoints.py
@@ -90,6 +90,17 @@ async def anthropic_response(  # noqa: PLR0915
             user_api_key_dict=user_api_key_dict, data=data, call_type="text_completion"
         )
 
+        tasks = []
+        tasks.append(
+            proxy_logging_obj.during_call_hook(
+                data=data,
+                user_api_key_dict=user_api_key_dict,
+                call_type=ProxyBaseLLMRequestProcessing._get_pre_call_type(
+                    route_type="anthropic_messages"  # type: ignore
+                ),
+            )
+        )
+
         ### ROUTE THE REQUESTs ###
         router_model_names = llm_router.model_names if llm_router is not None else []
 
@@ -97,23 +108,21 @@ async def anthropic_response(  # noqa: PLR0915
         if (
             llm_router is not None and data["model"] in router_model_names
         ):  # model in router model list
-            llm_response = asyncio.create_task(llm_router.aanthropic_messages(**data))
+            llm_coro = llm_router.aanthropic_messages(**data)
         elif (
             llm_router is not None
             and llm_router.model_group_alias is not None
             and data["model"] in llm_router.model_group_alias
         ):  # model set in model_group_alias
-            llm_response = asyncio.create_task(llm_router.aanthropic_messages(**data))
+            llm_coro = llm_router.aanthropic_messages(**data)
         elif (
             llm_router is not None and data["model"] in llm_router.deployment_names
         ):  # model in router deployments, calling a specific deployment on the router
-            llm_response = asyncio.create_task(
-                llm_router.aanthropic_messages(**data, specific_deployment=True)
-            )
+            llm_coro = llm_router.aanthropic_messages(**data, specific_deployment=True)
         elif (
             llm_router is not None and data["model"] in llm_router.get_model_ids()
         ):  # model in router model list
-            llm_response = asyncio.create_task(llm_router.aanthropic_messages(**data))
+            llm_coro = llm_router.aanthropic_messages(**data)
         elif (
             llm_router is not None
             and data["model"] not in router_model_names
@@ -122,9 +131,9 @@ async def anthropic_response(  # noqa: PLR0915
                 or len(llm_router.pattern_router.patterns) > 0
             )
         ):  # model in router deployments, calling a specific deployment on the router
-            llm_response = asyncio.create_task(llm_router.aanthropic_messages(**data))
+            llm_coro = llm_router.aanthropic_messages(**data)
         elif user_model is not None:  # `litellm --model <your-model-name>`
-            llm_response = asyncio.create_task(litellm.anthropic_messages(**data))
+            llm_coro = litellm.anthropic_messages(**data)
         else:
             raise HTTPException(
                 status_code=status.HTTP_400_BAD_REQUEST,
@@ -134,8 +143,16 @@ async def anthropic_response(  # noqa: PLR0915
                 },
             )
 
-        # Await the llm_response task
-        response = await llm_response
+        tasks.append(llm_coro)
+
+        # wait for call to end
+        llm_responses = asyncio.gather(
+            *tasks
+        )  # run the moderation check in parallel to the actual llm api call
+
+        responses = await llm_responses
+
+        response = responses[1]
 
         hidden_params = getattr(response, "_hidden_params", {}) or {}
         model_id = hidden_params.get("model_id", None) or ""
@@ -183,6 +200,11 @@ async def anthropic_response(  # noqa: PLR0915
                 headers=dict(fastapi_response.headers),
             )
 
+        ### CALL HOOKS ### - modify outgoing data
+        response = await proxy_logging_obj.post_call_success_hook(
+            data=data, user_api_key_dict=user_api_key_dict, response=response # type: ignore
+        )
+
         verbose_proxy_logger.info("\nResponse from Litellm:\n{}".format(response))
         return response
     except Exception as e:
diff --git a/tests/test_litellm/integrations/test_custom_guardrail.py b/tests/test_litellm/integrations/test_custom_guardrail.py
index e71b68ab93..182e013492 100644
--- a/tests/test_litellm/integrations/test_custom_guardrail.py
+++ b/tests/test_litellm/integrations/test_custom_guardrail.py
@@ -1,4 +1,4 @@
-from unittest.mock import AsyncMock, MagicMock
+from unittest.mock import AsyncMock
 
 import pytest
 
@@ -82,3 +82,101 @@ class TestCustomGuardrailDeploymentHook:
         # Verify messages were updated in result
         assert result["messages"] == mock_result["messages"]
         assert result["messages"] != original_messages
+
+
+class TestCustomGuardrailShouldRunGuardrail:
+
+    def test_should_run_guardrail_with_litellm_metadata(self):
+        """Test that should_run_guardrail works with litellm_metadata pattern"""
+        from litellm.types.guardrails import GuardrailEventHooks
+        
+        custom_guardrail = CustomGuardrail(
+            guardrail_name="test_guardrail", 
+            default_on=False,
+            event_hook=GuardrailEventHooks.pre_call
+        )
+        
+        # Test with guardrails in litellm_metadata
+        data = {
+            "model": "gpt-3.5-turbo",
+            "litellm_metadata": {
+                "guardrails": ["test_guardrail"]
+            }
+        }
+        
+        result = custom_guardrail.should_run_guardrail(
+            data=data, event_type=GuardrailEventHooks.pre_call
+        )
+        
+        assert result is True
+        
+    def test_should_run_guardrail_with_metadata(self):
+        """Test that should_run_guardrail works with metadata pattern"""
+        from litellm.types.guardrails import GuardrailEventHooks
+        
+        custom_guardrail = CustomGuardrail(
+            guardrail_name="test_guardrail", 
+            default_on=False,
+            event_hook=GuardrailEventHooks.pre_call
+        )
+        
+        # Test with guardrails in metadata
+        data = {
+            "model": "gpt-3.5-turbo",
+            "metadata": {
+                "guardrails": ["test_guardrail"]
+            }
+        }
+        
+        result = custom_guardrail.should_run_guardrail(
+            data=data, event_type=GuardrailEventHooks.pre_call
+        )
+        
+        assert result is True
+        
+    def test_should_run_guardrail_with_root_level_guardrails(self):
+        """Test that should_run_guardrail works with root level guardrails"""
+        from litellm.types.guardrails import GuardrailEventHooks
+        
+        custom_guardrail = CustomGuardrail(
+            guardrail_name="test_guardrail", 
+            default_on=False,
+            event_hook=GuardrailEventHooks.pre_call
+        )
+        
+        # Test with guardrails at root level
+        data = {
+            "model": "gpt-3.5-turbo",
+            "guardrails": ["test_guardrail"]
+        }
+        
+        result = custom_guardrail.should_run_guardrail(
+            data=data, event_type=GuardrailEventHooks.pre_call
+        )
+        
+        assert result is True
+        
+        
+    def test_should_run_guardrail_no_matching_guardrail(self):
+        """Test that should_run_guardrail returns False when guardrail name doesn't match"""
+        from litellm.types.guardrails import GuardrailEventHooks
+        
+        custom_guardrail = CustomGuardrail(
+            guardrail_name="test_guardrail", 
+            default_on=False,
+            event_hook=GuardrailEventHooks.pre_call
+        )
+        
+        # Test with different guardrail name
+        data = {
+            "model": "gpt-3.5-turbo",
+            "litellm_metadata": {
+                "guardrails": ["different_guardrail"]
+            }
+        }
+        
+        result = custom_guardrail.should_run_guardrail(
+            data=data, event_type=GuardrailEventHooks.pre_call
+        )
+        
+        assert result is False