diff --git a/examples/aio/multi_agent.py b/examples/aio/multi_agent.py
new file mode 100644
index 0000000..cf2116a
--- /dev/null
+++ b/examples/aio/multi_agent.py
@@ -0,0 +1,80 @@
+import asyncio
+from typing import Sequence
+
+from absl import app, flags
+
+import xai_sdk
+from xai_sdk.chat import user
+from xai_sdk.tools import web_search, x_search
+
+THINKING = flags.DEFINE_bool("thinking", True, "Whether the model display the reasoning tokens consumed or not")
+
+
+async def basic_response(chat: xai_sdk.aio.chat.Chat):
+    """Sample and print a complete multi-agent response."""
+    # Sample a response from a multi-agent model.
+    response = await chat.sample()
+
+    print(f"Grok: {response.content}")
+
+    # `cost_usd` is per-request; it shows us  the response costs in us dollars.
+    if response.cost_usd is not None:
+        print(f"Cost: ${response.cost_usd:.4f}")
+
+
+async def response_with_thinking(chat: xai_sdk.aio.chat.Chat):
+    """Stream a Multi-agent response with reasoning progress."""
+
+    print("Grok: ", end="", flush=True)
+
+    content_started = False
+    last_response = None
+
+    async for response, chunk in chat.stream():
+        last_response = response
+
+        if response.usage.reasoning_tokens and not content_started:
+            print(f"\rThinking...({response.usage.reasoning_tokens} tokens)", end="", flush=True)
+
+        if chunk.content:
+            if not content_started:
+                print("\rGrok: ", end="", flush=True)
+                content_started = True
+
+            print(chunk.content, end="", flush=True)
+
+    print()
+
+    if last_response is None:
+        raise RuntimeError("The model returned no response.")
+
+    # `cost_usd` is per-request; it shows us  the response costs in us dollars.
+    if last_response.cost_usd is not None:
+        print(f"Cost: ${last_response.cost_usd:.4f}")
+
+
+async def main(argv: Sequence[str]) -> None:
+    if len(argv) > 1:
+        raise app.UsageError("Unexpected command line arguments.")
+
+    client = xai_sdk.AsyncClient()
+
+    chat = client.chat.create(
+        model="grok-4.20-multi-agent",
+        # you can choose 4 or 16 agent
+        agent_count=4,
+        # you can use tools such as web search, x search, or/and code execution
+        tools=[web_search(), x_search()],
+        include=["verbose_streaming"],
+    )
+
+    chat.append(user("Research the latest breakthroughs in quantum computing and summarize the key findings."))
+
+    if THINKING.value:
+        await response_with_thinking(chat)
+    else:
+        await basic_response(chat)
+
+
+if __name__ == "__main__":
+    app.run(lambda argv: asyncio.run(main(argv)))