MonteWalk

Instructions

Logs an agent action or reasoning step for audit purposes.

Args:
    action_type: Category (e.g., 'REASONING', 'TRADE_DECISION', 'ERROR').
    details: Description of the action.

Implementation Reference

• tools/logger.py:41-52 (handler)

  The core handler function for the 'log_action' tool. It takes action_type and details, logs them formatted to the logger, and returns a success message. The function signature and docstring provide the input schema.

  def log_action(action_type: str, details: str) -> str:
      """
      Logs an agent action or reasoning step for audit purposes.
      
      Args:
          action_type: Category (e.g., 'REASONING', 'TRADE_DECISION', 'ERROR').
          details: Description of the action.
      """
      # Clean up details to remove excessive newlines or emojis if needed
      clean_details = details.strip()
      logger.info(f"[{action_type.upper()}] {clean_details}")
      return "Action logged successfully."

• server.py:400-403 (registration)

  Registers the log_action function as an MCP tool by calling register_tools, which dynamically applies the @mcp.tool() decorator from FastMCP.

  register_tools(
      [log_action],
      "Logging"
  )

• server.py:355-364 (helper)

  Helper function used to register MCP tools, including log_action, by applying the FastMCP decorator and logging the process.

  def register_tools(tools: List[Callable], category: str):
      """Helper to register multiple tools with logging."""
      for tool in tools:
          try:
              mcp.tool()(tool)
              logger.info(f"Registered {category} tool: {tool.__name__}")
          except Exception as e:
              logger.error(f"Failed to register {tool.__name__}: {e}")
              raise

• app.py:312-313 (registration)

  Includes log_action in the tools_map dictionary used to generate Gradio interfaces for the tool, which are exposed via MCP when mcp_server=True.

      "Utils": [log_action],
  }