diff --git a/ai-hub/integration_tests/test_provider_config.py b/ai-hub/integration_tests/test_provider_config.py
new file mode 100644
index 0000000..bc54cff
--- /dev/null
+++ b/ai-hub/integration_tests/test_provider_config.py
@@ -0,0 +1,126 @@
+import os
+import httpx
+import pytest
+
+BASE_URL = os.getenv("SYNC_TEST_BASE_URL", "http://127.0.0.1:8002/api/v1")
+
+def _headers():
+    return {
+        "X-User-ID": os.environ.get("SYNC_TEST_USER_ID", "")
+    }
+
+def test_verify_llm_success_gemini():
+    """
+    Test 1: Check LLM verification with a valid API Key and Model format directly.
+    """
+    payload = {
+        "provider_name": "gemini",
+        "api_key": os.getenv("GEMINI_API_KEY", "fallback_key_if_missing"),
+        "model": "gemini/gemini-3-flash-preview"
+    }
+    
+    with httpx.Client(timeout=15.0) as client:
+        r = client.post(f"{BASE_URL}/users/me/config/verify_llm", headers=_headers(), json=payload)
+        
+        assert r.status_code == 200, f"Expected 200 OK: {r.text}"
+        data = r.json()
+        
+        # Depending on if the key works for litellm.acompletion('Hello') right now
+        # If it returns success: True or False doesn't matter for the HTTP connection, 
+        # but logically it should be processed without 400s or 500s.
+        # If the key provided by the user is real, 'success' will be True.
+        assert "success" in data, "Response missing 'success'"
+
+def test_verify_llm_failure_invalid_key():
+    """
+    Test 2: Check LLM verification with an explicitly invalid API Key fails gracefully.
+    """
+    payload = {
+        "provider_name": "gemini",
+        "api_key": "junk_invalid_key_123",
+        "model": "gemini/gemini-3-flash-preview"
+    }
+    
+    with httpx.Client(timeout=15.0) as client:
+        r = client.post(f"{BASE_URL}/users/me/config/verify_llm", headers=_headers(), json=payload)
+        
+        assert r.status_code == 200, f"Expected 200 OK: {r.text}"
+        data = r.json()
+        assert data["success"] is False, "Verification should fail for an invalid key."
+        assert "message" in data and len(data["message"]) > 0
+
+def test_update_user_llm_preferences():
+    """
+    Test 3: Update LLM user preferences fully and ensure they save.
+    """
+    payload = {
+        "llm": {
+            "active_provider": "gemini",
+            "providers": {
+                "gemini": {
+                    "api_key": os.getenv("GEMINI_API_KEY", "fallback_key_if_missing"),
+                    "model": "gemini/gemini-3-flash-preview"
+                }
+            }
+        },
+        "tts": {},
+        "stt": {},
+        "statuses": {}
+    }
+    
+    with httpx.Client(timeout=10.0) as client:
+        r = client.put(f"{BASE_URL}/users/me/config", headers=_headers(), json=payload)
+        assert r.status_code == 200, f"Expected 200 OK: {r.text}"
+
+        # Fetch config to verify
+        r_get = client.get(f"{BASE_URL}/users/me/config", headers=_headers())
+        assert r_get.status_code == 200
+        saved_prefs = r_get.json()
+        
+        # Verify provider configs
+        assert "preferences" in saved_prefs
+        assert "llm" in saved_prefs["preferences"]
+        assert saved_prefs["preferences"]["llm"]["active_provider"] == "gemini"
+        
+        # The API likely masks the key on GET requests inside `merge_user_config`
+        # We ensure it exists
+        saved_gemini = saved_prefs["preferences"]["llm"]["providers"].get("gemini", {})
+        assert "api_key" in saved_gemini
+        assert "model" in saved_gemini
+        assert saved_gemini["model"] == "gemini/gemini-3-flash-preview"
+
+def test_verify_llm_success_gemini_masked_key_fallback():
+    """
+    Test 4: Verify that a masked key (***) correctly pulls the true key from 
+    the user's DB configuration automatically.
+    """
+    payload = {
+        "provider_name": "gemini",
+        "api_key": "***",
+        "model": "gemini/gemini-3-flash-preview"
+    }
+    
+    with httpx.Client(timeout=15.0) as client:
+        r = client.post(f"{BASE_URL}/users/me/config/verify_llm", headers=_headers(), json=payload)
+        
+        # If DB lookup succeeds, it executes test using the real key saved in Test 3
+        assert r.status_code == 200, f"Expected 200 OK: {r.text}"
+        data = r.json()
+        assert "success" in data
+
+def test_verify_llm_unrecognized_provider():
+    """
+    Test 5: Check behavior for a completely unrecognized provider name.
+    """
+    payload = {
+        "provider_name": "non_existent_provider_xyz",
+        "api_key": "123",
+        "model": "unknown_model"
+    }
+    with httpx.Client(timeout=10.0) as client:
+        r = client.post(f"{BASE_URL}/users/me/config/verify_llm", headers=_headers(), json=payload)
+        
+        assert r.status_code == 200
+        data = r.json()
+        # Should gracefully fail
+        assert data["success"] is False