Streaming agent response

streaming-agent-response

Use Runner.run_streamed() to get a RunResultStreaming object. Iterate result.stream_events() to receive raw token deltas or higher-level run items as they arrive.

When to apply

Chat UIs needing progressive text display, long-running agents where the user should see progress, or any scenario where latency-to-first-token matters.

Token-level streaming

import asyncio
from openai.types.responses import ResponseTextDeltaEvent
from agents import Agent, Runner

async def main():
    agent = Agent(name="Joker", instructions="You are a helpful assistant.")
    result = Runner.run_streamed(agent, input="Please tell me 5 jokes.")
    async for event in result.stream_events():
        if event.type == "raw_response_event" and isinstance(event.data, ResponseTextDeltaEvent):
            print(event.data.delta, end="", flush=True)

asyncio.run(main())

Item-level streaming

from agents import Agent, ItemHelpers, Runner

async def main():
    agent = Agent(name="Joker", instructions="First call how_many_jokes, then tell jokes.", tools=[...])
    result = Runner.run_streamed(agent, input="Hello")
    async for event in result.stream_events():
        if event.type == "raw_response_event":
            continue
        elif event.type == "run_item_stream_event":
            if event.item.type == "tool_call_item":
                print(f"Tool called: {event.item.raw_item.name}")
            elif event.item.type == "message_output_item":
                print(f"Message: {ItemHelpers.text_message_output(event.item)}")

Key notes

• Always drain stream_events() to completion; the run is not complete until the iterator ends
• RunResultStreaming.interruptions is populated after the stream ends if tool approval is needed
• Call result.cancel(mode="after_turn") to stop cleanly after the current turn
• result.is_complete reflects final run state only after the iterator finishes