diff --git a/letta/agents/letta_agent.py b/letta/agents/letta_agent.py
index 49a50cb1..a5b4d4ec 100644
--- a/letta/agents/letta_agent.py
+++ b/letta/agents/letta_agent.py
@@ -1021,6 +1021,7 @@ class LettaAgent(BaseAgent):
                             is_openai_proxy=agent_state.llm_config.provider_name == "lmstudio_openai",
                             messages=current_in_context_messages + new_in_context_messages,
                             tools=request_data.get("tools", []),
+                            put_inner_thoughts_in_kwarg=agent_state.llm_config.put_inner_thoughts_in_kwargs,
                         )
                     else:
                         raise ValueError(f"Streaming not supported for {agent_state.llm_config}")
diff --git a/letta/interfaces/openai_streaming_interface.py b/letta/interfaces/openai_streaming_interface.py
index 08f23e0f..8a4d968f 100644
--- a/letta/interfaces/openai_streaming_interface.py
+++ b/letta/interfaces/openai_streaming_interface.py
@@ -35,13 +35,14 @@ class OpenAIStreamingInterface:
         is_openai_proxy: bool = False,
         messages: Optional[list] = None,
         tools: Optional[list] = None,
+        put_inner_thoughts_in_kwarg: bool = True,
     ):
         self.use_assistant_message = use_assistant_message
         self.assistant_message_tool_name = DEFAULT_MESSAGE_TOOL
         self.assistant_message_tool_kwarg = DEFAULT_MESSAGE_TOOL_KWARG
 
         self.optimistic_json_parser: OptimisticJSONParser = OptimisticJSONParser()
-        self.function_args_reader = JSONInnerThoughtsExtractor(wait_for_first_key=True)  # TODO: pass in kwarg
+        self.function_args_reader = JSONInnerThoughtsExtractor(wait_for_first_key=put_inner_thoughts_in_kwarg)
         self.function_name_buffer = None
         self.function_args_buffer = None
         self.function_id_buffer = None