Agent as tool does not keep the run config

[jack-lac]

run_config is not passed to the new run when we use agent.as_tool(). Therefore, if a custom provider is set, it cannot run the tool agent.

Debug information

• Agents SDK version: 0.0.9
• Python version: 3.11.11

Expected behavior

run_config must be passed in a way to the runner of "as_tool agent" so the "tool-run" is done with the same configuration than the main run.

openai-agents-python/src/agents/agent.py

Line 211 in 84fb734

output = await Runner.run(

[rm-openai]

Hmm good point. I'm not sure if we should pass through the entire config, since it may have settings we don't want to apply to the sub-agent. There's also a workaround - you can just have your own @function_tool that passes the context along.

Perhaps we could just document it - or maybe we ask for an optional config to be passed in. Thoughts?

[rm-openai]

#504 has docs, which I think is better than trying to make as_tool comprehensive.