FAQ

Is CanalAPI really OpenAI-compatible?

Yes. The base path is /v1 and the request / response shape matches OpenAI's HTTP API for the supported endpoints (most importantly Chat Completions). Most OpenAI client libraries work as drop-in clients once the baseURL and apiKey are overridden.

How do I find the current model list?

The console Models page is the authoritative source. Model names and prices can change as upstream providers release new versions.

How is usage billed?

CanalAPI uses prepaid balance. Each request deducts the corresponding token cost from your account balance. See Billing for details.

Can I use CanalAPI from the browser?

You should not embed your API key in client-side JavaScript or in a mobile app — that exposes the key to anyone who can read the bundle. Use a server-side proxy (for example a Next.js Route Handler) and call CanalAPI from your own backend.

Does CanalAPI store my prompts?

CanalAPI's user-facing usage logs record metadata (token counts, model, timestamps, status), not full prompts or completions. Refer to the privacy policy in the console for the authoritative description of what is collected and retained.

What about rate limits?

See Rate Limits. Honor Retry-After when present and use exponential backoff with jitter otherwise.

Can I get an invoice for my company?

Yes. Invoices are issued by the operating Japanese entity. Configure your invoice header (company name, registration number, address) in Billing → Profile before topping up.

Where can I report a bug?

Use the support form in the console. Include the request ID (returned in response headers when available) and the approximate timestamp.