Get started

One key for every model

1

Create a key

One key works for every model. It is shown once, so store it as KORPI_API_KEY.

Create key
2

Add funds

One balance pays for video, image and text. Top up from $10; failed generations are refunded.

Add funds
3

Send a request

Call any model by id. Chat is OpenAI compatible, so existing SDKs work with a new base URL.

Playground

Build a request

Model
Grok Imaginegrok-imagine
Duration
Aspect ratio

Estimated cost

$0.0594

videos/generations
curl -X POST https://backend.korpi.ai/v1/videos/generations \
  -H "Authorization: Bearer $KORPI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "grok-imagine",
    "prompt": "Macro shot of a white paper origami crane slowly unfolding its wings on a dark walnut desk, warm desk lamp light, floating dust particles, smooth slow camera push-in, cinematic",
    "duration": 6,
    "aspect_ratio": "16:9"
  }'
 
# Poll until status is completed
curl https://backend.korpi.ai/v1/videos/generations/$ID \
  -H "Authorization: Bearer $KORPI_API_KEY"
Sample response

Press Preview to see a real response.

Recipes

Every model, same key

POST /videos/generations
curl -X POST https://backend.korpi.ai/v1/videos/generations \
  -H "Authorization: Bearer $KORPI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "grok-imagine",
    "prompt": "Macro shot of a white paper origami crane slowly unfolding its wings on a dark walnut desk, warm desk lamp light, floating dust particles, smooth slow camera push-in, cinematic",
    "duration": 6,
    "aspect_ratio": "16:9"
  }'
 
# Poll until status is completed
curl https://backend.korpi.ai/v1/videos/generations/$ID \
  -H "Authorization: Bearer $KORPI_API_KEY"
OpenAI SDK

Keep your SDK

Point the base URL at Korpi and switch models by id.

https://backend.korpi.ai/v1
Chat completions
import os
from openai import OpenAI
 
client = OpenAI(base_url="https://backend.korpi.ai/v1", api_key=os.environ["KORPI_API_KEY"])
 
completion = client.chat.completions.create(
    model="claude-opus-5.5",
    messages=[{"role": "user", "content": "Hello"}],
)
print(completion.choices[0].message.content)
Agents

Hand it to your agent

Paste into Claude Code, Cursor or Codex
Use the Korpi API (base URL https://backend.korpi.ai/v1) with the key in the KORPI_API_KEY environment variable, sent as "Authorization: Bearer $KORPI_API_KEY".
- Video: POST /videos/generations with {"model":"grok-imagine","prompt":"...","duration":6|8|10,"aspect_ratio":"16:9"|"9:16"|"1:1"} and optional "image_url". It returns 202 with an id; poll GET /videos/generations/{id} every 3 s until status is "completed" or "failed".
- Image: POST /images/generations with {"model":"nano-banana","prompt":"...","aspect_ratio":"1:1"} and up to 4 https "image_urls" as references.
- Text: POST /chat/completions in OpenAI format. GET /models lists every model id.
- Errors are {"error":{"code","message"}}. On 402 insufficient_balance, stop and tell me to add funds.
Errors

Every error, explained

400invalid_requestA field is missing or out of range. The message names it.
400model_not_foundUnknown model id. Call GET /models for the list.
401invalid_api_keyMissing, malformed or revoked key.
402insufficient_balanceAdd funds and retry. Nothing was charged.
404not_foundUnknown endpoint or generation id.
503api_disabledTemporarily unavailable. Retry with backoff.