## Log of AI calls: feature, model, tokens, latency, outcome – never prompts or answers (desk:admin, KI-016)

`GET /api/desk/v1/ai/calls`

Module: Desk · operationId: `listAiCalls` · Tag: AI · Auth: access token (`Authorization: Bearer <accessToken>`)

### Parameters

| Name | In | Type | Required | Description |
|---|---|---|---|---|
| `feature` | query | string |  | (one of `triage`, `subject`, `summary`, `draftReply`) |
| `limit` | query | integer |  | Rows per page (default 50) (default `50`; max 200) |
| `cursor` | query | string |  | The `nextCursor` of the previous page (keyset, bound to the list and its order); anything else is 400 invalid with `field: cursor`. |

### Responses

- **200** Calls (`application/json`)

Fields of the 200 answer:

- `items` (array of object, required)
- `nextCursor` (string): Cursor of the next page (pass it as `cursor`); missing on the last page.

Example 200 answer:

```json
{
  "items": [
    {
      "actorRef": "sophie.huber@muster-it.example",
      "confidence": 0.91,
      "createdAt": "2026-11-17T07:12:20Z",
      "customerId": "3c5e7a9c-1d3f-4b5a-9c7e-1a3c5e7a9c11",
      "feature": "triage",
      "id": "4b0c1d2e-5f6a-4b7c-9d8e-0f1a2b3c4d02",
      "inputTokens": 1840,
      "itemId": "4a1b2c3d-5e6f-4a7b-8c9d-1e2f3a4b5c01",
      "latencyMs": 1210,
      "model": "decision",
      "outcome": "succeeded",
      "outputTokens": 96,
      "providerKind": "platform",
      "role": "decision"
    }
  ],
  "nextCursor": "eyJpZCI6IjRiMGMxZDJlIn0"
}
```

### Example request

```bash
curl -X GET 'https://api.mbcom.cloud/api/desk/v1/ai/calls' \
  -H "Authorization: Bearer $ACCESS_TOKEN"
```
