Providers
Providers
Which providers the gateway reaches, which you can connect with your own key, and how to list your own service.
One Experiential key reaches models across many providers. Some rungs are platform-funded (paid with credits), some run on your own provider account, and inference providers can list their own endpoints on the marketplace.
Coverage
| Provider | Platform-funded rungs | Your own key (BYOK) |
|---|---|---|
| OpenAI | Yes | Yes |
| Anthropic | Yes | Yes |
| Google Gemini and Vertex AI | Yes | Yes |
| Azure OpenAI (including Azure AI Foundry) | Yes | Yes |
| Amazon Bedrock | Yes | Yes |
| OpenRouter | Yes | Yes |
| Fireworks | Yes | Yes |
| Experiential Cloud | Yes | Not applicable |
| DeepSeek, Mistral, xAI, Z.ai, Tencent, Meta and others | Per model | Yes |
| Modal | No | Yes, your deployment |
| Local OpenAI-compatible server | No | Yes, a private model |
Claude on Azure is served only on your own key. The live list of platform-funded providers is the providers map on the catalog.
Other providers you can connect with your own key: Qwen, DashScope, Cerebras, Huawei, Wafer and TypeSafe. Your own free accounts at Groq, NVIDIA NIM, Kenari, KiosAPI and TokenHarbor connect on every plan, without Pro.
/api/modelscurl -s "https://api.experientiallabs.ai/api/models?limit=1000" | jq '.providers | keys'Becoming a provider
If you run a public OpenAI-compatible HTTPS endpoint, you can list it. The gateway routes customer traffic to you, bills at the rates you publish and shows your lane in the catalog with measured latency, throughput and uptime.
| You control | |
|---|---|
| Price | Nano-USD per million tokens, per model; increases take effect no sooner than 7 days after you request them |
| Capacity | Declare limits; return 429 any time and traffic falls through to another provider |
| Lifecycle | Add, deprecate and remove models |
| Data policy | A required declaration: retention, training, ZDR. Customers filter on it |
Apply from your organization's Inference page. A human reviews every application. New lanes start at the tail of a model's chain and move up on observed performance. Earnings accrue to a provider wallet, net of an agreed platform fee.
/api/orgs/{org_id}/provider/applyApply
In-app application, with no credentials in it.Review and key exchange
We store your API key in a vault through a private channel.Canary, then live
Your first lane passes a smoke check, then serves real traffic.