Python MCP Client takes time more than 30s to find tools on Java Spring AI MCP Server

[ashwinbhite]

Pre-submission Checklist

• I have verified that this discussion would not be more appropriate as an issue in a specific repository
• I have searched existing discussions to avoid duplicates

Discussion Topic

I have mcp python client which connects to java spring ai mcp server having registered @tools as "search_workspaces".
Issue : When the client sends a prompt to the server, the connection is established quickly. However, there is a significant delay—around 30 seconds—before the system identifies and selects the "search_workspaces" tool. This latency in identification of tool leads to an overall slow response time.

python client code

         connection_kwargs = {
        "autocommit": True,
        "prepare_threshold": 0 ,
        "row_factory": dict_row
    }
    
    model = ChatOpenAI(model="gpt-4o-mini", openai_api_key=OPENAI_API_KEY)

   def generate_workspace_system_prompt() -> str:
      preferences  = """
      You are about to access the workspaces 
      tokenString is {token}
      default pageno 1 and pagesize 20
      default sort is updatedTimestamp:desc
      orgId is {orgId}
      Returns name as workspaceName, workspaceId, type and accountName a list of my roles.    
      """
      return preferences
     async with MultiServerMCPClient(server_config) as client:

              all_tools = client.get_tools();
              tools_map = {tool.name: tool for tool in all_tools}
              print(f"tool name ::: {tools_map}")
   
   search_workspaces_agent = create_react_agent(
                  model= model,
                  tools= [tools_map.get("search_workspaces")],
                  name= "search_workspaces",
                  prompt= generate_workspace_system_prompt().format(token=token, orgId=orgId)
              );
   workflow_agent = create_supervisor(
                  [search_workspaces_agent],
                  model=model,
                  add_handoff_back_messages=False,
                  output_mode='last_message',
                  # tools= all_tools,
                  prompt=generate_system_prompt()
              )
              agent_response = []
              async with AsyncConnectionPool(
                      conninfo=DB_URI,
                      max_size=20,
                      min_size=20,
                      open=True,
                      kwargs=connection_kwargs,
              ) as pool:
                  checkpointer = AsyncPostgresSaver(pool)
                  print(f"checkpointer  :: {checkpointer}")

            await checkpointer.setup()
            configuration = {"configurable": {"thread_id": id, "user_id": userId}}
            agent =  workflow_agent.compile(checkpointer= checkpointer)

            response = await agent.ainvoke({"messages": input_query}, config= configuration)