Developer preview

Build with Mavis AI.

A familiar interface for your next application. Here’s how the planned integration fits into your existing code.

API access is not live yet

These examples describe the planned launch interface. Model requests, file uploads, and key creation will be enabled when the gateway launches.

1. Create an account

Your workspace for models and usage.

2. Add credits

Prepaid billing opens at launch.

3. Create an API key

Keep the key on your application’s server.

Base URL

Set your OpenAI client’s base URL to the address below. Authenticate with a Mavis API key using the standard Bearer header.

https://api.mavis.app/v1

Your first streaming request

Install the OpenAI SDK and set MAVIS_API_KEY in your server environment. The examples below use the Responses API.

from openai import OpenAI
import os

client = OpenAI(
    base_url="https://api.mavis.app/v1",
    api_key=os.environ["MAVIS_API_KEY"],
)

stream = client.responses.create(
    model="gpt-6.1-sol",
    input="What will we build today?",
    stream=True,
)

for event in stream:
    if event.type == "response.output_text.delta":
        print(event.delta, end="", flush=True)

Planned endpoints

Method	Path	Purpose
POST	/v1/responses	Responses and streaming
POST	/v1/chat/completions	Chat Completions compatibility
GET	/v1/models	Available model catalog
POST	/v1/files	Upload supported files

Streaming, images, and files

The gateway will preserve supported OpenAI request formats and streaming events. Image and document support depends on the model and endpoint. Files uploaded to another provider account won’t automatically be available in your Mavis workspace.

A feature matrix, upload limits, billing rules, and error reference will be published with the API release.

Explore the model catalog