FAQ

Can I use one key for both OpenAI and Anthropic?

Yes. The same API key and balance work for /v1/chat/completions and /v1/messages. Which models you can call over each protocol depends on how each model's endpoints are configured (see API access).

Is the balance shared across models?

Yes — one balance (in USD) is used by every model. Each model has its own price, so the spend rate differs.

How is caching billed?

Cached input is billed separately and far cheaper than normal input. The exact cache read/write tokens and the charge are recorded for every request on the Usage page.

Which models are available to me?

The models in your group. See the Models & pricing page or call GET /v1/models with your key — it lists exactly what your account can use, with the supported protocols.

Why does a model work in one client but not another?

Most likely a protocol mismatch. A model is only reachable over the protocol of its configured endpoint, and we never auto-convert. Use the protocol the model supports (OpenAI vs Anthropic) — see Troubleshooting.

What happens when my balance runs out?

New API requests stop with a visible insufficient_balance error. Your cabinet, keys and history stay available; top up to resume.

Do you store my prompts/chats?

We store usage metadata (model, tokens, cache, charge, status) for billing and your logs — not your message content for training.

Are there rate limits?

Per-request limits and the number of API keys are governed by your group. Finer policies (rate limits, IP allowlists, schedules) are on the roadmap.