CrewAI

CrewAI reaches this API through its LLM class with custom_openai=True, which sends calls through the OpenAI SDK to your base_url. Give the LLM to each agent with llm=.

Install

pip install "crewai[openai]"

uv add "crewai[openai]"

Configure API keys

.env

API_KEY=sk-...

Load it into the environment before you run the code (for example with python-dotenv). Keep the key out of your code.

Run

Text

import os
from crewai import LLM

llm = LLM(
    model="openai/gpt-6-luna",
    custom_openai=True,
    base_url="https://pinerouter.ai/v1",
    api_key=os.environ["API_KEY"],
)

print(llm.call("Name three rivers in Europe."))

Agents and tools

from crewai import Agent, Crew, Task
from crewai.tools import tool

@tool("get_weather")
def get_weather(city: str) -> str:
    """Current weather for a city."""
    return f"Sunny, 21 C in {city}"

forecaster = Agent(
    role="Forecaster",
    goal="Answer weather questions",
    backstory="You check the weather tool before you answer.",
    tools=[get_weather],
    llm=llm,
)
task = Task(description="What is the weather in {city}?", expected_output="One sentence.", agent=forecaster)

print(Crew(agents=[forecaster], tasks=[task]).kickoff(inputs={"city": "Paris"}))

Stream

Add stream=True to the LLM(...) call to stream the model's answers.

Specifics

custom_openai=True needs a base_url (or OPENAI_BASE_URL in the environment). With it, CrewAI drops the openai/ prefix and sends gpt-6-luna to us.
Without custom_openai=True, CrewAI checks openai/... names against OpenAI's own model list and sends unknown ids through LiteLLM instead.
For images, video and other media call Jobs with the HTTP helper inside a tool.
Tool calling depends on the model. Test with the model you plan to use.