---
title: "API billing and privacy"
url: "https://stylome.io/api/docs/billing-and-privacy"
---

> [Documentation index](https://stylome.io/llms.txt) · [API overview](https://stylome.io/api/docs)

# API billing and privacy

Both public models cost **$5 per million normalized Unicode code points**. API access uses prepaid Stripe credit and is separate from personal extension subscriptions. Review the [privacy policy](https://stylome.io/privacy) and [terms](https://stylome.io/terms) before sending content.

## What is charged

The billable count is the normalized input length used for inference and input limits. A successful 1,000-character request costs `$0.005000`. A successful 10,000-character request costs `$0.050000`. These are arithmetic examples, not recorded scoring calls.

Successful calls, including cache hits, are charged. Failed calls are free. An idempotent replay returns the original operation's usage without a second debit. A fresh successful request without replay is charged again, even if it contains the same text.

Scoring returns `usage.input_characters`, exact six-decimal `usage.cost_usd`, `usage.currency: "USD"`, and `usage.pricing_version`. Use decimal arithmetic for money. Do not infer charges from a floating-point model score or token estimate.

Fund API credit through your [account](https://stylome.io/account). The configured top-up range is exposed by `GET /api/v1/limits`; the approved default range is $5–$1,000. Requests without enough available credit return `402 insufficient_credit`. Account billing review can prevent scoring. The limits endpoint shows balance, reserved credit, available credit, and review status.

## Retention modes

Choose the policy when commissioning a key. It is fixed for that key. Both `stylome-1.2-large` and `stylome-1.0-large` support `zero_content`; existing keys use standard retention unless issued otherwise.

| Mode | Behavior |
| --- | --- |
| `standard` | Supports retained scoring records, cache entries, receipts where applicable, and idempotent replay |
| `zero_content` | Does not retain submitted text, excerpts, content hashes, classifier output, receipts, cache entries, or idempotency records |

Zero-content responses still return the result and content hash to the caller. Stylome does not store them. Your own client and logging systems can retain data independently; configure them for your privacy needs.

Zero-content keys retain operational metadata: key retention mode, request ID, timestamp, model and scoring contract, HTTP status, latency, normalized character count, USD charge, and rate-limit state. Character counts and costs are billing metadata only. Zero customer-content retention does not mean zero operational metadata.

## Replay and privacy

Replay requires stored request-derived data. Zero-content keys therefore reject `Idempotency-Key` with `400 idempotency_not_supported`. They bypass scoring content caches and do not create receipts. An application cannot obtain replay by omitting or changing a per-request privacy option.

Standard-retention keys can replay completed requests for 24 hours. Follow the [idempotency rules](https://stylome.io/api/docs/operations). Treat receipt share tokens as private capabilities and avoid logging or forwarding them unintentionally.

## Integration checklist

Before making a scoring call, select the required retention mode, verify the credential at `GET /api/v1/key`, and inspect quota and credit at `GET /api/v1/limits` with `account:read`. Keep Bearer secrets on your server. Redact secrets and customer text from application logs. Use request IDs for support.

The API is in early alpha. There is no guaranteed notice period for endpoint changes or model retirement. Read [scoring](https://stylome.io/api/docs/scoring) for interpretation limits; a classification is not proof of authorship or misconduct.