API pricing

Fees, usage rates and request budgets for integrations. For costs in terms of your writing, estimate how many texts your credits cover. Live pricing data.

Credits and charges

Usage prices include service overhead and payment-processing costs. No amount is deducted from a credit purchase before tax. Tax, if due, is added by checkout and does not become credits. Credits have no scheduled expiry, cannot be transferred, and do not trigger renewals. For unused-credit refunds, contact the service owner through Nathan's website. Payment-processing fees may not be recoverable; applicable consumer rights still apply.

A successful call uses the provider's reported input and output token counts, including billed reasoning tokens. Cached input receives no separate discount in this first price schedule. Each call rounds up to the next millionth of a dollar. Text is never sent to an unpriced model on a prepaid account.

Choose Low for routine edits, Medium for everyday writing, High for demanding edits, or Max for the highest capability. Medium is the default. Higher levels can cost more and take longer. Models may change within each level.

Use model_level with low, medium, high or max in AI requests and POST /v1/billing/quote. Use the same level for the quote and request. Your response and balance history show the actual charge.

Your spending limit

Each paid request supplies X-Max-Cost-USD, from $0.00575 to $10, and an Idempotency-Key. Unslop holds the full budget first. Before each model call, it checks that the remaining budget covers a conservative input/output estimate. It returns unused funds when the request finishes. Parallel requests cannot spend the same credits.

If a request fails, you pay nothing. We return the full held amount, including fees for any model calls already attempted. Rewrites with no accepted edits are also free. Successful results use reported token counts; if those counts are unavailable, the reserved estimate applies.

If the next call would exceed your cap, it is not sent. The API returns an error and refunds the entire request. Retrying an identical key retrieves the prior outcome; it does not run the job again. If the server stops before finishing, its hold expires after ten minutes. Your next balance check or processing request returns those credits. Retrying the interrupted request returns a saved failure without charging or running it again.

Payment and text records

Stripe hosts checkout and handles card details. Unslop grants credits only after verifying a signed Stripe webhook and checking the payment with Stripe. Duplicate events do not grant extra credit. Refunds remove the corresponding credits; a negative balance or unresolved dispute blocks paid use.

To support retries, prepaid requests may cache their response body in private account storage. A response can contain your draft, excerpts or revision. Up to five responses are retained for up to one hour and removed on the next paid request after expiry, so idle accounts can retain them longer. Larger responses are not cached. Charge records, request hashes and financial records remain; submitted request bodies are not stored. Saved voice profiles keep the text you chose to save.

Deleting an account requires a zero balance and no pending holds. Payment records remain after deletion for reconciliation. Provider and hosting retention policies still apply.

What usage charges cover

Usage charges cover AI processing, hosting, storage and payment fees. Card fees vary by country and card. Existing owner-managed keys are self-funded and bypass prepaid charging.