Frequently asked developer questions
Everything about connecting, architecture, billing, and privacy
Why is this cheaper than wiring up each AI API myself?
You get a single balance, a single API key, and one OpenAI-compatible request format for Claude, GPT and Gemini. No minimum deposit at each maker separately, no currency conversions.
Do you store my prompts or sensitive data?
Request bodies are never written to logs. Supported native Responses models store responses by default, as at OpenAI; send store: false to avoid storing the response. If you supply a conversation, its input and output items are still saved. Bridged models do not store responses. Stored responses and conversations are encrypted: a response for 30 days, a conversation until you delete it, and you can delete either at any time. Local files you upload to /v1/files are kept until you delete them, and at most 30 days after upload. Two more things are worth knowing rather than discovering later. When a provider refuses or fails a request we keep the first 400 characters of its error, and a refusal on content grounds sometimes repeats the phrase it objected to. And the provider you called does receive the request itself, because it cannot answer without it; the privacy policy links what each provider does with it. Local files expire 30 days after upload; the returned file includes expires_at. You can have up to 1,000 live conversations per account. Creating a stored file, conversation or Responses API response with store: true requires a positive spendable balance; with $0 spendable, the request returns HTTP 402. Closing your account erases stored files, conversations and responses.
How does Smart Fallback work?
If a provider has an outage or answers with an error, the gateway retries the request automatically. You see one answer, not a failure.
Is your gateway compatible with Cursor, Cline, OpenHands, and LangChain?
Yes. Our gateway speaks the OpenAI chat format (`/v1/chat/completions`), so all you need to do is point the `baseURL` at `https://api.flintbeam.com/v1` and pick any model. Code written against the older `https://flintbeam.com/v1` keeps working. Images, speech in both directions and embeddings have their own endpoints too: `/v1/images/generations`, `/v1/audio/transcriptions`, `/v1/audio/speech` and `/v1/embeddings`. The Responses API (`/v1/responses`) works as well, so the Codex CLI does too. There is no fine-tuning.
What payment methods do you support?
Bank cards through Stripe — Visa, Mastercard, UnionPay and Apple Pay — Alipay, WeChat Pay, UPI, Pix and cryptocurrency: USDT (TRC-20 and ERC-20), USDC, BTC and ETH. All are credited automatically, and the exact list of coins and networks is the one shown at checkout.