diff --git a/tradingagents/default_config.py b/tradingagents/default_config.py
index f553ff8e..9c900100 100644
--- a/tradingagents/default_config.py
+++ b/tradingagents/default_config.py
@@ -17,7 +17,7 @@ DEFAULT_CONFIG = {
     "llm_provider": os.environ.get("LLM_PROVIDER", "openai"),
     "deep_think_llm": os.environ.get("DEEP_THINK_LLM", "gpt-5.2"),
     "quick_think_llm": os.environ.get("QUICK_THINK_LLM", "gpt-5-mini"),
-    "backend_url": os.environ.get("BACKEND_URL", os.environ.get("LLAMACPP_BASE_URL", "https://api.openai.com/v1")),
+    "backend_url": os.environ.get("BACKEND_URL", "https://api.openai.com/v1"),
     # Provider-specific thinking configuration
     "google_thinking_level": None,      # "high", "minimal", etc.
     "openai_reasoning_effort": None,    # "medium", "high", "low"
diff --git a/tradingagents/llm_clients/openai_client.py b/tradingagents/llm_clients/openai_client.py
index da16f3a9..f7e74434 100644
--- a/tradingagents/llm_clients/openai_client.py
+++ b/tradingagents/llm_clients/openai_client.py
@@ -57,7 +57,7 @@ class OpenAIClient(BaseLLMClient):
             llm_kwargs["base_url"] = "http://localhost:11434/v1"
             llm_kwargs["api_key"] = "ollama"  # Ollama doesn't require auth
         elif self.provider == "llamacpp":
-            base_url = self.base_url or os.environ.get("LLAMACPP_BASE_URL", "http://localhost:8080/v1")
+            base_url = os.environ.get("BACKEND_URL") or os.environ.get("LLAMACPP_BASE_URL", "http://localhost:8080/v1")
             llm_kwargs["base_url"] = base_url
             llm_kwargs["api_key"] = "no-key-needed"  # llama-server doesn't require auth
         elif self.base_url: