Trust center

Clear answers before you send your first request

This page summarizes operational behavior. The Privacy Policy and Terms of Service remain the controlling documents.

Request data

For standard text API calls, operational logs store usage metadata such as model, token counts, status, duration, quota use, and timestamps. They are not designed to store complete prompts or generated text. Requests are transmitted to the selected model provider to produce a response.

Model routing

The requested public model name is mapped to an enabled upstream channel. Retry and failover can select another configured channel for that model after eligible upstream failures. The Models page is the source of truth for models currently offered to customers.

Billing

One prepaid Credit represents one USD of API value. Model-specific input and output rates are shown before use. Usage is settled from measured token consumption; failed requests that produce no billable model usage are not presented as successful calls.

Keys and access

API keys can be restricted by quota, model, status, and expiry. Secret credentials, Authorization headers, passwords, and full API keys are excluded from application logs.

Reliability

The gateway uses bounded retries, channel health controls, and streaming cleanup. No availability percentage is promised unless it is stated in a separate written agreement. Current reachability is published on the Status page.

Support and refunds

Questions about routing, usage, privacy, or billing can be sent to support@your-api.xyz. Refund eligibility is governed by the published Refund Policy.