First Agent

Minimal example

import asyncio

from agiwo.agent import Agent, AgentConfig
from agiwo.llm import OpenAIModel


async def main() -> None:
    agent = Agent(
        AgentConfig(
            name="assistant",
            description="Helpful SDK assistant",
            system_prompt="Answer concisely and use tools only when needed.",
        ),
        model=OpenAIModel(name="gpt-5.4"),
    )

    result = await agent.run("What makes a runtime harness useful for AI agents?")
    print(result.response)

    await agent.close()


asyncio.run(main())

Run it with:

uv run python hello.py

Stream the response

async for event in agent.run_stream("Explain recursion in one sentence."):
    if event.type == "step_delta" and event.delta.content:
        print(event.delta.content, end="", flush=True)

If you need full control over waiting or cancellation, use start() and then consume handle.stream().

Add tools

Builtin tools are available by default, including bash, bash_process, web_search, web_reader, and memory_retrieval.

You can also register your own BaseTool implementations:

agent = Agent(
    AgentConfig(
        name="math-assistant",
        description="Uses tools for arithmetic",
        system_prompt="Use the calculator tool for arithmetic.",
    ),
    model=OpenAIModel(name="gpt-5.4"),
    tools=[CalculatorTool()],
)

When to move beyond a single agent

Stay with Agent.run() and run_stream() when:

a single run is enough
you do not need persistent runtime state
you do not need child-agent orchestration

Move to the scheduler when you need:

long-lived roots
waiting and wake conditions
multi-agent coordination
explicit steering or cancellation

First Agent

First Agent

Minimal example

Stream the response

Add tools

When to move beyond a single agent

Next steps