FLUX 3 API & MCP

Using Claude, ChatGPT, Cursor or another AI app? You don't need code. Paste this into your AI and it connects itself:

Copy

Connect FLUX 3 to this chat so I can make images and videos here. Setup: https://flux3.video/connect.md

Step-by-step for each app

• Who can use it: accounts that have made a purchase on FLUX 3.
• Keys and connected apps: Account → API & MCP.
• What it covers: text to image, image to image, text to video and image to video, with the models this site offers. The model list can change; read it from GET /api/v1/models.

Quickstart

1. Create an API key on your account page. It is shown once, so store it like a password.
2. Start a generation:

export API_KEY="bnt_your_key"

curl -X POST https://flux3.video/api/v1/generations \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"flux3-image","capability_type":"t2i","prompt":"a paper fox in fresh snow"}'

3. Fetch the result with the id from the answer. When status is succeeded, outputs[0].url is your image; it is also in your assets on FLUX 3.

curl https://flux3.video/api/v1/generations/GENERATION_ID -H "Authorization: Bearer $API_KEY"

The examples use FLUX 3 Klein (flux3-image). Any model from GET /api/v1/models works the same way.

Using an AI app instead? You do not need a key: add the MCP server below and sign in.

Connect an AI app (MCP)

MCP server URL:

https://flux3.video/mcp

It is a remote MCP server (Streamable HTTP) with OAuth sign-in. Add the URL to your AI app; the app opens FLUX 3 so you can sign in and approve it. Only accounts that have made a purchase can approve. You can disconnect an app, or change its daily credit limit, on your account page.

Tools

generate waits up to about 25 seconds. If the result is ready it returns the output URLs; otherwise it returns the generation id, and get_generation returns the result once it is done.

Things to ask your AI app:

• "What image models does FLUX 3 have, and what do they cost?"
• "Make four square product shots of a ceramic mug on a wooden table."
• "Turn this photo into a short video" (with an https:// link to the photo)
• "How many credits do I have left today?"

REST API

Base URL: https://flux3.video/api/v1

Send your API key in the Authorization header. Keys only work on FLUX 3.

curl https://flux3.video/api/v1/balance \
  -H "Authorization: Bearer $API_KEY"

Endpoints

Generate an image, step by step

1. Pick a model and read its parameters:

curl https://flux3.video/api/v1/models -H "Authorization: Bearer $API_KEY"
curl https://flux3.video/api/v1/models/flux3-image -H "Authorization: Bearer $API_KEY"

2. Optionally check the price first with POST /api/v1/quote (same body, nothing is charged).

3. Start the generation. It answers 202 right away:

curl -X POST https://flux3.video/api/v1/generations \
  -H "Authorization: Bearer $API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: my-first-image" \
  -d '{"model":"flux3-image","capability_type":"t2i","prompt":"a paper fox in fresh snow"}'

{
  "success": true,
  "data": {
    "id": "3f9a1c2b7d4e8f60",
    "status": "queued",
    "model": "flux3-image",
    "capability_type": "t2i",
    "type": "image",
    "prompt": "a paper fox in fresh snow",
    "params": {
      "prompt": "a paper fox in fresh snow"
    },
    "outputs": [],
    "credits_quoted": 10,
    "error": null,
    "created_at": "2026-10-20T08:00:00.000Z"
  }
}

4. Poll GET /api/v1/generations/{id} every few seconds until status is succeeded or failed. Statuses go queued → running → succeeded | failed; images take seconds, videos a few minutes.

{
  "success": true,
  "data": {
    "id": "3f9a1c2b7d4e8f60",
    "status": "succeeded",
    "model": "flux3-image",
    "capability_type": "t2i",
    "type": "image",
    "prompt": "a paper fox in fresh snow",
    "params": {
      "prompt": "a paper fox in fresh snow"
    },
    "outputs": [
      {
        "type": "image",
        "url": "https://flux3.video/api/cdn/file/generated/example.webp",
        "mime": "image/webp",
        "width": 1024,
        "height": 1024,
        "duration_s": null
      }
    ],
    "credits_quoted": 10,
    "error": null,
    "created_at": "2026-10-20T08:00:00.000Z"
  }
}

The ids, credits and file address above are examples. For image to image or image to video, put public https:// URLs of your images in reference_urls. To refuse a price higher than you expect, send max_credits: the call fails with PRICE_CHANGED instead of charging more.

Code examples

Python (with requests)

import os, time, requests

API = "https://flux3.video/api/v1"
HEADERS = {"Authorization": f"Bearer {os.environ['API_KEY']}"}

res = requests.post(
    f"{API}/generations",
    headers={**HEADERS, "Idempotency-Key": "paper-fox-1"},
    json={"model": "flux3-image", "capability_type": "t2i", "prompt": "a paper fox in fresh snow"},
).json()
if not res["success"]:
    raise SystemExit(res["error"]["code"])

gen = res["data"]
while gen["status"] in ("queued", "running"):
    time.sleep(3)
    gen = requests.get(f"{API}/generations/{gen['id']}", headers=HEADERS).json()["data"]

print(gen["status"], [out["url"] for out in gen["outputs"]])

JavaScript (Node 18+ or any runtime with fetch)

const API = "https://flux3.video/api/v1";
const headers = { Authorization: `Bearer ${process.env.API_KEY}` };

let res = await fetch(`${API}/generations`, {
  method: "POST",
  headers: { ...headers, "Content-Type": "application/json", "Idempotency-Key": "paper-fox-1" },
  body: JSON.stringify({ model: "flux3-image", capability_type: "t2i", prompt: "a paper fox in fresh snow" }),
}).then((r) => r.json());
if (!res.success) throw new Error(res.error.code);

let gen = res.data;
while (gen.status === "queued" || gen.status === "running") {
  await new Promise((done) => setTimeout(done, 3000));
  gen = (await fetch(`${API}/generations/${gen.id}`, { headers }).then((r) => r.json())).data;
}

console.log(gen.status, gen.outputs.map((out) => out.url));

Responses and errors

Every response is JSON and carries an X-Request-Id header.

{ "success": true, "data": { "...": "..." } }
{ "success": false, "error": { "code": "INSUFFICIENT_CREDITS", "message": "...", "request_id": "..." } }

Branch on error.code, never on the message. The ones you are most likely to meet:

All codes: VALIDATION_ERROR, INVALID_PARAMS, INVALID_MODEL, MODERATION_BLOCKED, UNAUTHORIZED, INVALID_TOKEN, TOKEN_EXPIRED, PAYMENT_REQUIRED, USER_BANNED, INSUFFICIENT_CREDITS, PREMIUM_CREDITS_REQUIRED, DAILY_CAP_REACHED, MODEL_NOT_AVAILABLE, NOT_FOUND, BATCH_NOT_FOUND, PRICE_CHANGED, IDEMPOTENCY_MISMATCH, RATE_LIMITED, INTERNAL_ERROR, SERVICE_OVERLOADED, SERVICE_UNAVAILABLE.

Credits and limits

• A generation costs the same credits as on the site. Credits are taken when the generation starts; failed outputs are refunded automatically.
• Daily credit limit: each API key and each connected app may spend up to 1,000 credits per UTC day by default. Change it or turn it off on your account page. Past the limit, new generations fail with DAILY_CAP_REACHED until 00:00 UTC.
• Rate limits: about 10 new generations and 120 other requests per minute per account; past that you get 429 RATE_LIMITED.
• Retries: send an Idempotency-Key header with POST /api/v1/generations. Repeating a request with the same key returns the first generation instead of starting (and charging for) a new one; reusing a key with a different body fails with IDEMPOTENCY_MISMATCH.

Machine-readable

• OpenAPI 3.1: https://flux3.video/developers/openapi.json
• This page as Markdown: https://flux3.video/developers.md