Requests and responses in a run

A run is a conversation between the agent and the model, kept as a list of messages you can print to see exactly what was sent and returned.

Last updated: 28 Sep, 2026 · Pydantic AI 2.51

The run in Agent basics and a first run returned an answer. That answer came from a message exchange, and result.all_messages() hands you the whole list.

Listing the messages of a run

for message in result.all_messages():
    print(message.kind)
    for part in message.parts:
        print("   ", part.part_kind)

request
    user-prompt
response
    text

What the two messages hold

• A request goes from the agent to the model; a response comes back.
• Each message holds parts. The request has one user-prompt part, your ticket; the response has one text part, the answer.
• kind and part_kind are the two labels you read to follow a run.

Watching a run grow with a tool

A tool is a Python function the model can ask the agent to call. Function tools in Pydantic AI covers tools in full; one here is enough to see the message list grow:

@agent.tool_plain
def lookup_order(order_id: str) -> str:
    """Look up an order's status."""
    return f"Order {order_id} has shipped."


result = agent.run_sync("Where is my order A-1001?")
for message in result.all_messages():
    for part in message.parts:
        content = getattr(part, "content", None) or getattr(part, "args", None)
        print(f"{message.kind:8} {part.part_kind:12} {content}")
print(result.usage)

request  user-prompt  Where is my order A-1001?
response tool-call    {'order_id': 'a'}
request  tool-return  Order a has shipped.
response text         {"lookup_order":"Order a has shipped."}
RunUsage(input_tokens=115, output_tokens=17, requests=2, tool_calls=1)

Reading the four messages

• The first response is a tool-call part with arguments, not the answer.
• The agent runs the tool and sends the result back as a tool-return part in a new request.
• The model then writes its text answer, so the run made two requests and usage shows tool_calls=1.
• The test model called the tool with order_id='a': it calls every tool once with made-up arguments that fit the types. A real model would read A-1001 from the ticket.

A request vs a response

When you read the message list

• Working out why an agent called a tool, or which arguments it sent.
• Storing a conversation to continue it later, covered in the message-history part.
• Debugging a run that made more requests than you expected.

Related

• Previous: Agent basics and a first run
• Next: FunctionModel: write a stand-in model
• Reference: Message history