This is a good question. I'll have to think of a long term good answer, but the short term hack is: just add the tools in the initiate_gmail_connection tool call.

my_agent = Agent(instructions=[...])

@function_tool
def initiate_gmail_connection():
   my_agent.tools.extend([send_gmail, read_gmail, ...])

Thanks @rm-openai , unfortunately, it seems that extending the tools isn't enough reason to refresh the current_span / agent_span in the _run_streamed_impl (

Any other ideas?

@Afstkla I have a system with dynamic agents. I mean users define their agents by entering instructions, settings handoffs, and assigning tools from a predefined set of tools. I fetch the agents config of the user when a request arrives to my backend and dynamically create agents and run them. This system works well so far.

The problem occurs where more than one tool need to be executed in order. Let me give me an example. Let's say we have two tools one for creating a discount code (create_discount) and one for sending an email (send_email). If they want to show it in the chat, they can only add create_dicount tool to the agent and write the prompt accordingly. This scenario works perfectly in my system. However, if the user wants to send the discount code to the customer's email then they add both tool to the agent and explain the flow in the prompt. In this scenario, the system only calls create_dicount but not send_email. I would expect from an agent to consider all tools again (like it does in the initial call) when it is called again with the tool response.

Is the any official way of doing what I desire? Or should we workaround this manually? Also if there is no official way, I think there should be because then developers can create strong atomic tools and users may achieve an infinite amount of different things using a this small set if tools. The current system don't allow this. What do you think @rm-openai?