Help
FAQ
Short answers to the questions people ask most about the gateway, billing and data.
Short answers, each with a link to the page that explains it.
Getting started
| Question | Answer |
|---|---|
| What is Experiential? | One OpenAI, Anthropic and Gemini compatible API for every model, with failover across providers, spend controls and data controls. |
| Do I need a new SDK? | No. Point the OpenAI or Anthropic SDK at https://api.experientiallabs.ai/v1 with your xpl_ key. See the Quickstart. |
| Which APIs are supported? | Chat Completions, Responses and Anthropic Messages, plus model listings and usage APIs. |
| Is there a markup on tokens? | No. Platform-funded tokens bill at the catalog price; plans add value through allotments and discounts. |
| Where do I find model ids? | GET /v1/models, or the catalog. Everything listed is callable. See Models and the catalog. |
curl https://api.experientiallabs.ai/v1/chat/completions \
-H "Authorization: Bearer YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{"model": "gpt-6-luna", "messages": [{"role": "user", "content": "Hello"}]}'Reliability
| Question | Answer |
|---|---|
| What happens if a provider is down? | The gateway fails over to the next rung before any output is sent. See Routing. |
| Can a response mix two providers? | No. Once output starts, it comes from one rung. |
| How do I see which provider answered? | The x-gateway-provider header and the body's provider field. |
| Why did I get a 429? | Your tier's window, too many queued requests, credits, or a free allowance. The code and message say which. See Rate limits. |
| Should my SDK retry too? | Set max_retries=0 if you let the gateway retry, or the two layers multiply. |
What does context_window include? | Prompt plus maximum output. Budget your prompt to max_input_tokens. |
Billing
| Question | Answer |
|---|---|
| What is a credit? | One cent of platform usage. See Credits and billing. |
| Why is a $1 hold on my card? | Card verification. It is released, not charged, and unlocks free and promotional models. |
| Do my own provider keys use credits? | No. Your provider bills you; the call reports usage.cost: 0 and is_byok: true. |
| What do Pro, Max and Ultra add? | Monthly credits, plan discounts, higher rate limits and Pro features. See Plans and pricing. |
| Can I stack a promo code with a plan bonus? | No. You get the single best bonus for each purchase. |
| How do I cap spend? | Per-key spend caps, budgets and spend alerts. |
Data and privacy
| Question | Answer |
|---|---|
| Do you store my prompts? | Only while the prompt capture switch is on, and never for requests routed under zero data retention. See Data controls. |
| Do you train on my data? | Content we do not store is never used. Stored content from Free plans is part of our Data Licensing Program (de-identified, may train models or be licensed); on paid plans only if you opt in. Turn storage off (Pro) to keep content out. See the privacy policy. |
| How do I get zero data retention? | Pro: the account-wide switch, the require_zdr policy, or provider: {"zdr": true} per request. |
Is x-gateway-zdr a signed guarantee? | No. It is our reading of the serving route's published terms. |
| Where are my provider keys kept? | In an encrypted vault; only the last four characters are ever shown. See Security. |
Free and providers
| Question | Answer |
|---|---|
What does :free on a model id mean? | The free lane only. It stops at the allowance and is never charged. See Free models. |
| Why do I need Pro to add my own key? | New BYOK connections are a Pro feature; organizations with a connection already keep it. See Bring your own key. |
| What is Experiential Cloud? | Serving capacity Experiential manages, shown as one provider. See Experiential Cloud. |
| Can I sell inference through you? | Yes. See Providers. |