OpenAI Agents SDK - TokenLab

Overview

Create an OpenAI-compatible TokenLab client and pass it to OpenAIChatCompletionsModel.

Environment

export TOKENLAB_API_KEY="sk-your-tokenlab-key"
export TOKENLAB_BASE_URL="https://api.tokenlab.sh/v1"
export TOKENLAB_MODEL="gpt-5.6-luna"

Example

import asyncio
import os

from openai import AsyncOpenAI

from agents import Agent, OpenAIChatCompletionsModel, Runner, set_tracing_disabled

client = AsyncOpenAI(
    base_url=os.getenv("TOKENLAB_BASE_URL", "https://api.tokenlab.sh/v1"),
    api_key=os.environ["TOKENLAB_API_KEY"],
)

set_tracing_disabled(disabled=True)


async def main():
    agent = Agent(
        name="Assistant",
        instructions="You are concise and practical.",
        model=OpenAIChatCompletionsModel(
            model=os.getenv("TOKENLAB_MODEL", "gpt-5.6-luna"),
            openai_client=client,
        ),
    )

    result = await Runner.run(agent, "Explain custom model providers in one paragraph.")
    print(result.final_output)


asyncio.run(main())

Scope

This path uses Chat Completions compatibility. TokenLab’s native Responses endpoint is available at https://api.tokenlab.sh/v1/responses, but SDK features that depend on OpenAI-specific Responses semantics should be verified with the exact model and workflow before production use.