Quickstart

Make your first request in a few minutes.

1. Create an API key

Open API keys in the console and choose Create new secret key. Give it a name, leave the permissions on All for now, and create it.

You only see the secret once

Copy it somewhere safe straight away. If you lose it, revoke it and create another.

2. Keep it in an environment variable

Your code should read the key from the environment, never contain it:

Terminal

export DECILLION_API_KEY="dk_live_…your key…"

3. Send a request

curl

JavaScript

Python

POST /v1/chat/completions

curl https://api.decillionai.com/v1/chat/completions \
  -H "Authorization: Bearer $DECILLION_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "google/gemini-3.8-flash",
    "messages": [
      {
        "role": "user",
        "content": "Say hello in one sentence."
      }
    ]
  }'

You get back the model's reply and what the request cost in US dollars:

Response

{
  "id": "chatcmpl_9f3a1c",
  "object": "chat.completion",
  "created": 1760000000,
  "model": "google/gemini-3.8-flash",
  "choices": [
    {
      "index": 0,
      "message": { "role": "assistant", "content": "Hello there, it's lovely to meet you!" },
      "finish_reason": "stop"
    }
  ],
  "usage": {
    "prompt_tokens": 14,
    "completion_tokens": 11,
    "total_tokens": 25,
    "cost": 0.0000186
  }
}

4. Stream the reply

Set stream to true and the reply arrives in pieces as the model writes it, which is what you want in a chat interface.

curl

JavaScript

Python

Streaming

curl https://api.decillionai.com/v1/chat/completions \
  -H "Authorization: Bearer $DECILLION_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "google/gemini-3.8-flash",
    "messages": [
      {
        "role": "user",
        "content": "Write a haiku about the sea."
      }
    ],
    "stream": true
  }'

Where next

•

Try prompts without writing code in the Playground, then use Get code.

•

Keep a request alive when a model is down with fallback models.

•

Generate images, audio and video: other model types.

•

Run code on a machine of its own with Computers.

•

Set a spend cap on the key, and read about limits.

Introduction

Authentication