Install Pydantic AI and run your first agent

An Agent holds a model and instructions. run_sync sends it a prompt and waits for the answer. The built-in test model runs it without an API key.

from pydantic_ai import Agent

agent = Agent("test", instructions="You answer customer support tickets.")
result = agent.run_sync("I was charged twice for one order")
print(result.output)
print(result.usage)

Agent("test", ...) picks Pydantic AI's test model. It never reads the prompt: it answers with a fixed text, which is why the output says success (no tool calls). It is built for tests, and here it lets you see an agent run before any real model is involved.

result.output is the answer. result.usage counts what the run cost: requests=1 because the agent called the model once. The test model does not count real tokens; it estimates them by counting words, so the numbers are plausible, not exact.

Three ways to run

• agent.run_sync(prompt) waits for the answer. Use it in scripts.
• await agent.run(prompt) is the async version, for web servers and anything that handles several tickets at once.
• agent.run_stream(prompt) gives you the answer piece by piece. That is lesson 16.

import asyncio


async def main():
    result = await agent.run("My parcel never arrived")
    print(result.output)


asyncio.run(main())

In a terminal

Save the first example as first_agent.py and run it:

python first_agent.py

The first time an agent runs in a process, Pydantic AI prints a banner with its version and what the agent has. It goes to stderr, not stdout, so piping the output into a file or another program leaves it out. Set PYDANTIC_AI_NO_BANNER=1 to turn it off. The outputs in the rest of this course leave it out.