Gateway API reference

Authentication

Send your API key as a Bearer token. Requests are billed to the key owner's credits. Get an API key

Authorization: Bearer sk_xxxxxxxxxxxxxxxx

Endpoints

Create a generation

Pass the model id, its unified input parameters and optionally a provider slug. The response returns immediately with the task id; image models on synchronous providers may already be succeeded.

curl https://01router.com/api/v1/generations \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "seedance-2.5",
    "provider": "fal",            // optional: omit to auto-route
    "input": {
      "prompt": "A red fox running through fresh snow, cinematic",
      "duration": 5,
      "resolution": "720p",
      "aspect_ratio": "16:9",
      "generate_audio": true
    }
  }'

{
  "id": "gen_6f1c…",
  "object": "generation",
  "model": "seedance-2.5",
  "type": "video",
  "provider": "kie",
  "status": "processing",        // queued | processing | succeeded | failed
  "input": { "prompt": "…", "duration": 5, "resolution": "720p", "aspect_ratio": "16:9", "generate_audio": true },
  "outputs": [],
  "error": null,
  "credits": 512,
  "price_usd": 2.0475,
  "created_at": "2026-10-23T08:00:00.000Z",
  "completed_at": null
}

Get the result

Poll every few seconds until status is succeeded or failed. Output URLs are provider-hosted and may expire — download them promptly.

curl https://01router.com/api/v1/generations/gen_6f1c… -H "Authorization: Bearer $API_KEY"

{ "id": "gen_6f1c…", "status": "succeeded", "outputs": ["https://…/video.mp4"], … }

Audio models

Music (Suno) and text-to-speech (Gemini TTS) models use the same /generations flow as images and video: create a task, then poll it. Outputs are audio file URLs.

curl https://01router.com/api/v1/generations \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gemini-3.8-flash-tts",
    "input": { "prompt": "Welcome back, let us get started.", "voice": "Kore", "style_instructions": "warm and upbeat" }
  }'

{ "id": "gen_9a2d…", "type": "audio", "status": "succeeded", "outputs": ["https://…/speech.wav"], … }

Chat completions

Chat models (GPT, Claude, Gemini, Grok, DeepSeek…) use the OpenAI Chat Completions format, so the official OpenAI SDKs work with base URL https://01router.com/api/v1. Set "stream": true for server-sent events. Credits for the prompt plus max_tokens (4,096 when unset) are held first, then settled to the real token usage once the reply ends; the same request id appears in the X-Gateway-Request-Id header and the task log.

curl https://01router.com/api/v1/chat/completions \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "claude-sonnet-5.5",
    "messages": [{ "role": "user", "content": "Write a haiku about routers" }],
    "max_tokens": 512,
    "stream": true
  }'

from openai import OpenAI

client = OpenAI(base_url="https://01router.com/api/v1", api_key="sk_…")
res = client.chat.completions.create(
    model="gpt-6-sol",
    messages=[{"role": "user", "content": "Hello!"}],
)
print(res.choices[0].message.content, res.usage)

Provider routing

Without provider we validate the input against every provider's supported values, price it, and submit to the cheapest one. If submission fails we refund and try the next provider (up to 3). With provider the request goes only there.

Quote

Same body as creating a generation. Returns every available provider sorted by price, plus the ones that cannot serve the request and why.

curl https://01router.com/api/v1/quote -H "Content-Type: application/json" \
  -d '{ "model": "nano-banana-pro", "input": { "prompt": "poster", "resolution": "2K" } }'

{
  "model": "nano-banana-pro",
  "providers": [
    { "provider": "kie", "provider_name": "KIE AI", "unit": "image", "quantity": 1, "price_usd": 0.117, "credits": 30 },
    { "provider": "openrouter", … }
  ],
  "unavailable": [{ "provider": "atlascloud", "reason": "resolution must be one of 1K, 2K" }]
}

Billing

Credits are charged when the task is submitted, based on the chosen provider's price for your parameters (per second of video, per image or per request). Failed or timed-out tasks are refunded automatically.

Errors

{ "error": { "message": "Insufficient credits: this request costs 512 credits", "code": "insufficient_credits" } }

All model ids and parameters: Models