# Usage and costs

Every call is recorded under the id its answer carried, with who served it, every attempt and what each cost. Read one call back, list your organization’s calls, or sum them by day.

## Tokens on each answer

Every answer reports the tokens it used: `usage` on a [`POST /models`](/docs/gateway/reference/chat/) reply and on the
`stop` frame of its stream, and OpenAI's `usage` on a [Chat Completions](/docs/gateway/reference/chat-completions/)
answer. Counts are the provider's own, not estimated by Momor.

## One call

Every answer carries the call's id in `X-Momor-Request-Id`, which is also the `id` of a Chat Completions answer. Read the
record back with the same API key, using [Get a generation](/docs/gateway/reference/generation/): it lists every
offering the call was put to, with each attempt's tokens and `costUsd`, what it cost at the price of the offering it ran
on.

## Your organization's calls

The organization's owner, signed in, lists its calls newest first:

```http
GET /organization/gateway/requests?from=2026-10-01T00:00:00Z&to=2026-10-22T00:00:00Z&skip=0&take=50
```

| Parameter | Notes |
|---|---|
| `from` | The earliest time a listed call was made at. |
| `to` | The time every listed call was made before. |
| `requestId` | The one call to list. |
| `skip` | How many calls to pass over. 0 by default. |
| `take` | Calls a page, from 1 to 200. 50 by default. |

The answer is `{ "count", "items" }`: how many calls match, and the page, each entry a call record.

## Usage by day

```http
GET /organization/gateway/usage?from=2026-10-01T00:00:00Z&to=2026-10-22T00:00:00Z
```

`from` and `to` are required. The answer sums the calls made in that range by UTC day, by who made them, by the
provider and model that served them, and by whether they ran on your own key, in that order, a page of rows at a time:
`skip` passes over that many rows (0 by default) and `take` sets how many a page holds, from 1 to 200 (50 by default).
The answer is `{ "count", "items" }`: how many rows the range sums to, and the page. Each row carries:

| Field | Notes |
|---|---|
| `day` | The UTC day. |
| `userId` | The API key, or the person signed in, that made the calls. |
| `provider` · `model` | Who served them. Absent for calls nothing served. |
| `ownKey` | Whether they ran on your own provider key. |
| `requests` · `failed` | How many calls, and how many did not end in `200`. |
| `inputTokens` · `outputTokens` · `cacheReadTokens` · `cacheWriteTokens` · `thoughtTokens` | The tokens their answers reported. |

What the calls cost in credits is on your organization's balance in Settings › Credits & billing. See
[credits and pricing](/docs/gateway/credits/).

## Data handling

The Gateway route is stateless: it does not create a Momor conversation or retain a prompt or response history for
later retrieval. Operational telemetry and the record of each call are metadata only: request id, key, what was asked
for, the selected model and provider, attempts, status, token counts, latency and attributed cost. They exclude
prompts, messages, tool arguments and results, files, parsed document text, provider reasoning and signatures, and
credentials.

That does not mean content stays inside Momor. The selected model provider receives the content needed to answer.
Automatic or directed [file conversion](/docs/gateway/files/) can also send file content or extracted content to a
model-backed converter. Retention, training and regional-processing guarantees depend on the selected provider and
credential contract. Search history and Workflow run and knowledge retention are separate product contracts.