Expose Internal Stream Items, Reasoning, and Tool Calls from Agents.asTools

[marinosabijan]

When using Agents.asTools, the parent agent only receives the child’s final output. It would be highly valuable if the parent could also access the child agent’s streamed reasoning, intermediate items, and tool calls. This would enable meta-agents, improve debugging and logging, and allow building UIs that display step-by-step execution, turning asTools into a transparent and controllable orchestration layer rather than a black box.

(I am aware that instead of using asTools we can get that functionality by manually creating agent runs in the tool calls, but this will make it more convenient if it's optional).

[seratch]

Thanks for the input. Indeed, this could be useful for some scenario. However, tools do not expose its internals by design, so I don't think this project can come up with a viable solution for this need.