diff --git a/vllm/entrypoints/openai/api_server.py b/vllm/entrypoints/openai/api_server.py
index 374e43fb15341..1e735da641df9 100644
--- a/vllm/entrypoints/openai/api_server.py
+++ b/vllm/entrypoints/openai/api_server.py
@@ -24,6 +24,7 @@ from fastapi import APIRouter, Depends, FastAPI, Form, HTTPException, Request
 from fastapi.exceptions import RequestValidationError
 from fastapi.middleware.cors import CORSMiddleware
 from fastapi.responses import JSONResponse, Response, StreamingResponse
+from starlette.concurrency import iterate_in_threadpool
 from starlette.datastructures import State
 from starlette.routing import Mount
 from typing_extensions import assert_never
@@ -846,6 +847,21 @@ def build_app(args: Namespace) -> FastAPI:
             response.headers["X-Request-Id"] = request_id
             return response
 
+    if envs.VLLM_DEBUG_LOG_API_SERVER_RESPONSE:
+        logger.warning("CAUTION: Enabling log response in the API Server. "
+                       "This can include sensitive information and should be "
+                       "avoided in production.")
+
+        @app.middleware("http")
+        async def log_response(request: Request, call_next):
+            response = await call_next(request)
+            response_body = [
+                section async for section in response.body_iterator
+            ]
+            response.body_iterator = iterate_in_threadpool(iter(response_body))
+            logger.info("response_body={%s}", response_body[0].decode())
+            return response
+
     for middleware in args.middleware:
         module_path, object_name = middleware.rsplit(".", 1)
         imported = getattr(importlib.import_module(module_path), object_name)
diff --git a/vllm/envs.py b/vllm/envs.py
index 23c304f124d36..e5025485a2501 100644
--- a/vllm/envs.py
+++ b/vllm/envs.py
@@ -270,6 +270,11 @@ environment_variables: dict[str, Callable[[], Any]] = {
     "VLLM_API_KEY":
     lambda: os.environ.get("VLLM_API_KEY", None),
 
+    # Whether to log responses from API Server for debugging
+    "VLLM_DEBUG_LOG_API_SERVER_RESPONSE":
+    lambda: os.environ.get("VLLM_DEBUG_LOG_API_SERVER_RESPONSE", "False").
+    lower() == "true",
+
     # S3 access information, used for tensorizer to load model from S3
     "S3_ACCESS_KEY_ID":
     lambda: os.environ.get("S3_ACCESS_KEY_ID", None),