# Quickstart

Run your first generation in a few minutes. All 1,496 models take the same kind of request and return the
same request object, for images, video and audio: learn it once here.

> [!NOTE]
> Base URL `https://api.doitong.com/v1`. Every call sends `Authorization: Bearer <your key>`, and the same key works for the HTTP
> API and the MCP server.

## 1. Get an API key

Create a key at [https://doitong.com/developers/dashboard#keys](https://doitong.com/developers/dashboard#keys) and keep it in an
environment variable: it is shown only once. Each key can carry its own budget, so you can hand one to an agent
without risking the whole account. `GET /account` returns your API balance.

```bash
export DOITONG_API_KEY="doitong_..."

curl "https://api.doitong.com/v1/account" \
  -H "Authorization: Bearer $DOITONG_API_KEY"
```
```javascript
const res = await fetch("https://api.doitong.com/v1/account", {
  headers: { Authorization: `Bearer ${process.env.DOITONG_API_KEY}` },
});

console.log(await res.json());
```
```python
import os
import requests

r = requests.get(
    "https://api.doitong.com/v1/account",
    headers={"Authorization": f"Bearer {os.environ['DOITONG_API_KEY']}"},
)
print(r.json())
```

Runs are paid from your account's **API balance** in API credits: 1,000 API credits = $1, and every model costs
exactly its maker's price. Top up at [https://doitong.com/developers/dashboard#balance](https://doitong.com/developers/dashboard#balance).

## 2. Quote, then run

`POST /quote/{model_id}` takes the input JSON and returns the exact price. It needs no key and costs nothing.
Running the same input costs exactly that: the quote is the charge. (A model charged by usage is the exception: its
quote is the most the run can cost, and the unused part is refunded after the run.)

```bash
curl -X POST "https://api.doitong.com/v1/quote/kling-video/v3/pro/image-to-video" \
  -H "Content-Type: application/json" \
  -d '{"image": "https://example.com/fox.jpg", "duration": 5, "generate_audio": true}'
```
```javascript
const res = await fetch("https://api.doitong.com/v1/quote/kling-video/v3/pro/image-to-video", {
  method: "POST",
  headers: { "Content-Type": "application/json" },
  body: JSON.stringify({ image: "https://example.com/fox.jpg", duration: 5, generate_audio: true }),
});

const quote = await res.json();
console.log(quote.credits, quote.usd); // 840 "0.84"
```
```python
import requests

r = requests.post(
    "https://api.doitong.com/v1/quote/kling-video/v3/pro/image-to-video",
    json={"image": "https://example.com/fox.jpg", "duration": 5, "generate_audio": True},
)
print(r.json()["credits"], r.json()["usd"])  # 840 0.84
```

```json 200 OK · quote
{
  "object": "quote",
  "model": "kling-video/v3/pro/image-to-video",
  "credits": 840,
  "usd": "0.84",
  "unit": "second",
  "quantity": 5,
  "unit_credits": 168,
  "mode": "exact",
  "breakdown": "5 s × 168 API credits"
}
```

Then send the same body to `POST /run/{model_id}`. Add `?wait=20` to hold the connection for up to 20 seconds: a fast
model returns the finished result in one round trip. Send an `Idempotency-Key` header so that a retry never runs (or
charges) twice.

```bash
curl -X POST "https://api.doitong.com/v1/run/flux/schnell?wait=20" \
  -H "Authorization: Bearer $DOITONG_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: lighthouse-001" \
  -d '{"prompt": "A lighthouse on a rocky coast at dusk, warm light in the window"}'
```
```javascript
const res = await fetch("https://api.doitong.com/v1/run/flux/schnell?wait=20", {
  method: "POST",
  headers: {
    Authorization: `Bearer ${process.env.DOITONG_API_KEY}`,
    "Content-Type": "application/json",
    "Idempotency-Key": "lighthouse-001",
  },
  body: JSON.stringify({ prompt: "A lighthouse on a rocky coast at dusk, warm light in the window" }),
});

const request = await res.json();
console.log(request.status, request.outputs[0]?.url);
```
```python
import os
import requests

r = requests.post(
    "https://api.doitong.com/v1/run/flux/schnell",
    params={"wait": 20},
    headers={
        "Authorization": f"Bearer {os.environ['DOITONG_API_KEY']}",
        "Idempotency-Key": "lighthouse-001",
    },
    json={"prompt": "A lighthouse on a rocky coast at dusk, warm light in the window"},
)
request = r.json()
print(request["status"], request["outputs"][0]["url"] if request["outputs"] else None)
```

The response is a request object:

```json 200 OK · request
{
  "id": "req_8k2m4q7x1z9c3v5b6n0p2r4t",
  "object": "request",
  "model": "flux/schnell",
  "status": "succeeded",
  "cost": { "credits": 3, "usd": "0.003", "charged": 3, "refunded": 0 },
  "outputs": [
    { "type": "image", "url": "https://content.doitong.com/dev-api/2026/10/20/req_8k2m4q7x1z9c3v5b6n0p2r4t/0-output.jpeg",
      "content_type": "image/jpeg", "width": 1024, "height": 768, "stored": true }
  ],
  "error": null
}
```

## 3. Follow the status

If the model is still working when the wait ends, you get `202 Accepted` with the request's `id` and its current
`status`. Every request moves through the same lifecycle:

Lifecycle: `queued` → `running` → `succeeded` | `failed` | `canceled` (final; failed and canceled are refunded).

Read the request again with its id. `?wait=20` works here too, so a simple loop is enough:

```bash
curl "https://api.doitong.com/v1/requests/req_8k2m4q7x1z9c3v5b6n0p2r4t?wait=20" \
  -H "Authorization: Bearer $DOITONG_API_KEY"
```
```javascript
const headers = { Authorization: `Bearer ${process.env.DOITONG_API_KEY}` };
let request = await (await fetch("https://api.doitong.com/v1/requests/req_8k2m4q7x1z9c3v5b6n0p2r4t?wait=20", { headers })).json();

while (request.status === "queued" || request.status === "running") {
  const res = await fetch(`https://api.doitong.com/v1/requests/${request.id}?wait=20`, { headers });
  request = await res.json();
}

console.log(request.status, request.outputs[0]?.url);
```
```python
import os
import requests

headers = {"Authorization": f"Bearer {os.environ['DOITONG_API_KEY']}"}
request = {"id": "req_8k2m4q7x1z9c3v5b6n0p2r4t", "status": "queued"}

while request["status"] in ("queued", "running"):
    request = requests.get(
        f"https://api.doitong.com/v1/requests/{request['id']}",
        params={"wait": 20},
        headers=headers,
    ).json()

print(request["status"], request["outputs"][0]["url"] if request["outputs"] else None)
```

When `status` is `succeeded`, `outputs[].url` holds links to the files on our storage. If the request `failed` or was
`canceled`, its API credits are refunded automatically and `error` says why. Cancel anything still `queued` or
`running`:

```bash
curl -X POST "https://api.doitong.com/v1/requests/req_8k2m4q7x1z9c3v5b6n0p2r4t/cancel" \
  -H "Authorization: Bearer $DOITONG_API_KEY"
```

For long video jobs you can skip polling and pass a `webhook_url` instead (see [Webhooks](https://doitong.com/developers/docs/webhooks)).

## 4. Files

Pass any public URL as an input, or ask `POST /files` for an upload URL, `PUT` the file there and use the returned
`file_url` (see [Files](https://doitong.com/developers/docs/files)). A succeeded request lists its results in `outputs[]` with
type, content type and, when known, dimensions and duration.

```bash
curl -X POST "https://api.doitong.com/v1/files" \
  -H "Authorization: Bearer $DOITONG_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"filename": "fox.jpg", "content_type": "image/jpeg", "size": 482113}'
```

> [!TIP]
> **Files kept forever, or 7 days for free: your choice.** Output files are copied to our
> storage before a request succeeds. By default they stay until you delete them (100 API credits
> per GB per month); switch "Keep files forever" off in the dashboard, or pass `?storage=temporary`, to keep them
> 7 days for free (see [Files](https://doitong.com/developers/docs/files#storage)).

## Errors

Every model fails the same way: one envelope with a `type`, an exact `code`, the offending `param` and every problem
in `details`, so an agent can fix the request without reading prose. Nothing is charged for an invalid request.

```json 422 Unprocessable Entity
{
  "error": {
    "type": "invalid_request",
    "code": "invalid_input",
    "message": "duration must be one of 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15.",
    "param": "duration",
    "details": [ { "param": "duration", "message": "duration must be one of 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15." } ],
    "request_id": null,
    "doc_url": "https://doitong.com/developers/docs/errors#invalid_input"
  }
}
```

Every type and code is listed in [Errors](https://doitong.com/developers/docs/errors).

## Pricing

Prices are in USD and API credits and shown on every model page before you run. Each request carries `cost` with
`credits`, `usd`, `charged` and `refunded`. A run is charged when it is accepted; `failed` and `canceled` requests are
refunded automatically. Some models are priced exactly from the input, some by the length or size of the files you
send, others charge an *estimated price per run*: either way the quote is what you pay. A few models (language and
vision models, voice, processing-time tools) are *charged by usage*: the quote is the most the run can cost, and the
part the run did not use is refunded automatically. See [Pricing](https://doitong.com/developers/docs/pricing).

## MCP and AI agents

Connect any MCP client to `https://api.doitong.com/v1/mcp` with the same Bearer key. Your agent can search the catalog, quote and
run models, and read every docs page as Markdown.

```bash Claude Code
claude mcp add --transport http doitong-models https://api.doitong.com/v1/mcp \
  --header "Authorization: Bearer YOUR_API_KEY"
```

## Next steps

- Browse [all models](https://doitong.com/developers/models). Each page has a request builder that writes the code for you.
- Learn the request lifecycle: [Requests](https://doitong.com/developers/docs/requests).
- Use the same parameter names on every model: [Parameters](https://doitong.com/developers/docs/parameters).
- Working with an AI coding agent? Give it [https://doitong.com/developers/llms.txt](https://doitong.com/developers/llms.txt) or connect the [MCP server](https://doitong.com/developers/docs/mcp).