Skip to main content
This is not available yet. This page describes what is coming, so you can plan for it. Until then, use what the page suggests below.

What it will do

You will be able to point the official OpenAI SDK at your agent’s base URL and call responses.create or chat.completions.create. Your agent receives the request, including model and every other field, and its reply comes back in the OpenAI format, streamed or not.

What to use today

Call the invoke API directly: Python and TypeScript each take a few lines with the standard HTTP client.
Last modified on September 28, 2026