Documentation

Quickstart
Authentication
Models
Chat Completions
Errors
Rate limits
SDKs
Data & retention

MarsAI is a unified, OpenAI-compatible API for AI model inference. One endpoint and one key give you frontier and open-source models, served on our own GPU infrastructure.

If you've used the OpenAI API before, you already know ours. Point your existing client at the base URL below, swap in your MarsAI key, and every request, response, and streaming format behaves exactly as you expect.

Base URL

https://api.marsai.com/v1

What you get

• One API for frontier and open models, with automatic failover
• Drop-in OpenAI compatibility — no rewrites required
• Usage-based pricing with no markup on premium models

Quickstart

Make your first request in under five minutes.

1. 1

   Get an API key

   Request access and we'll email a key scoped to your organization.

2. 2

   Point your client at the base URL

   Set base_url to https://api.marsai.com/v1 in any OpenAI-compatible client.

3. 3

   Send a request

   Call the chat completions endpoint and read the response — stream it or take it whole.

Request

curl https://api.marsai.com/v1/chat/completions \
  -H "Authorization: Bearer $MarsAI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "claude-opus-4-8",
    "messages": [{"role": "user", "content": "Hello"}]
  }'

Response

{
  "id": "chatcmpl-9a1b...",
  "object": "chat.completion",
  "model": "claude-opus-4-8",
  "choices": [
    {
      "index": 0,
      "message": { "role": "assistant", "content": "Hello! How can I help?" },
      "finish_reason": "stop"
    }
  ],
  "usage": { "prompt_tokens": 9, "completion_tokens": 7, "total_tokens": 16 }
}

Authentication

All requests are authenticated with a bearer token sent in the Authorization header.

Your key carries the full permissions of your account, so keep it secret. Store it in an environment variable instead of hard-coding it, and never expose it in client-side code.

Example header

Authorization: Bearer $MarsAI_API_KEY

SDKs

Because the API is OpenAI-compatible, the official OpenAI SDKs work today — only the base URL changes.

Python

Install the OpenAI Python package, then set base_url and api_key.

Install

pip install openai

from openai import OpenAI

client = OpenAI(
    base_url="https://api.marsai.com/v1",
    api_key="$MarsAI_API_KEY",
)

resp = client.chat.completions.create(
    model="claude-opus-4-8",
    messages=[{"role": "user", "content": "Hello"}],
)
print(resp.choices[0].message.content)

JavaScript / TypeScript Coming soon

The OpenAI JS SDK already works against our base URL. A native MarsAI SDK is on the way.

import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://api.marsai.com/v1",
  apiKey: process.env.MarsAI_API_KEY,
});

const resp = await client.chat.completions.create({
  model: "claude-opus-4-8",
  messages: [{ role: "user", content: "Hello" }],
});
console.log(resp.choices[0].message.content);

Examples

Common requests across the chat, image, and embedding endpoints.

curl https://api.marsai.com/v1/chat/completions \
  -H "Authorization: Bearer $MarsAI_API_KEY" \
  -d '{
    "model": "claude-opus-4-8",
    "messages": [{"role": "user", "content": "Write a haiku about GPUs"}],
    "stream": true
  }'

Error codes

The API uses standard HTTP status codes. These are the ones you'll encounter most.