EP 03 / 234:56Live now

Give Your Agent a Tool

Name: Give Your Agent a Tool
Description: Turn any Python function into an agent capability with the @tool decorator.

Turn any Python function into an agent capability with the @tool decorator.

Agents that only talk are demos. Agents that use tools are systems.

In AG2, any Python function becomes a tool by adding the @tool decorator. The agent reads the function's docstring + type hints to learn when and how to call it.

What we build: a weather agent that calls a real get_weather(city) function when asked. Same pattern works for database queries, file reads, API calls, anything Python can do.

Watch for: the docstring is the prompt the LLM sees describing the tool. Bad docstring → bad tool use. Spend time on it.

Code

from autogen import ConversableAgent, tool

@tool
def get_weather(city: str) -> str:
    """Return the current weather for the given city.
    Args:
      city: city name, e.g. 'San Francisco'
    """
    # In real life, call a weather API here.
    return f"It's 18°C and clear in {city}."

travel = ConversableAgent(
    name="travel",
    llm_config={"model": "gpt-4o"},
    tools=[get_weather],
)

travel.run("What should I pack for Lisbon tomorrow?")

Give Your Agent a Tool

Links

Code