Your own keys
Bring your own provider keys
Add your own OpenAI, Anthropic, Google, AWS, Azure or other keys. Requests on them are free on DigitSpace's side; the provider bills you.
Add a key you already have with a provider, and DigitSpace sends that provider's requests on it. The provider bills you at your own rates and discounts; DigitSpace charges nothing for those requests. Routing, failover, budgets, guardrails and the request log work the same as on credits.
Add a key
Owners and admins add keys under Provider keys in the dashboard. Choose the provider, paste the key, and fill in anything else it needs (a region, a resource name). The key is checked with the provider before it's saved.
Give each key a description so you can tell them apart. Keys are stored encrypted, and only one active key per provider is used. To limit what runs on your keys, use budgets and access rules on your API keys; they count requests on your provider keys too.
Which requests use it
Each project chooses under Settings → Preferences → How requests are paid:
- Own keys: only your keys. A model whose providers you have no key for returns
503 no_provider_available. - Own keys, then credits (the default): your key for a provider you have one for, credits for the rest.
- Credits: never your keys.
Models the catalogue doesn't list
For most providers, adding a key also reads the list of models the key can use. You can then call those models even if DigitSpace's catalogue doesn't list them, by name or as provider/model, and they run on your key. They're listed by GET /v1/models when you send your key. Use Sync models in the key's menu to read the list again after the provider adds models.
AWS Bedrock, Google Vertex AI and Azure don't give a usable model list, so nothing is read from them. Their keys are used for the models DigitSpace's catalogue routes to that provider; see the provider's page under Providers. On the Enterprise plan you can add other deployments yourself as custom models.
Providers
| Provider | ID in provider/model |
|---|---|
| Alibaba Cloud | alibaba |
| Anthropic | anthropic |
| Atlas Cloud | atlascloud |
| AWS Bedrock | aws-bedrock |
| AWS Bedrock (OpenAI-compatible) | aws-mantle |
| Azure OpenAI | azure |
| Azure AI Foundry | azure-ai-foundry |
| Claude on Azure | azure-anthropic |
| Baidu Qianfan | baidu |
| ByteDance (BytePlus ModelArk) | bytedance |
| CanopyWave | canopywave |
| Cerebras | cerebras |
| DeepInfra | deepinfra |
| DeepSeek | deepseek |
| EmberCloud | embercloud |
| Fireworks AI | fireworks |
| Google AI Studio | google-ai-studio |
| Google Vertex AI (Gemini) | google-vertex |
| Google Vertex AI (partner models) | vertex-openai |
| Claude on Google Vertex AI | vertex-anthropic |
| Groq | groq |
| Inference.net | inference.net |
| Meta Model API | meta |
| MiniMax | minimax |
| Mistral AI | mistral |
| Moonshot AI | moonshot |
| NanoGPT | nanogpt |
| Nebius AI Studio | nebius |
| Novita AI | novita |
| OpenAI | openai |
| Perplexity | perplexity |
| RunPod | runpod |
| Sakana AI | sakana |
| SCX.ai | scx-ai |
| Tencent Hunyuan (TokenHub) | tencent |
| Together AI | together-ai |
| xAI | xai |
| Xiaomi MiMo | xiaomi |
| Z.ai | zai |
| Custom (OpenAI-compatible) | custom |
Guides for the providers that need more than a key:
For OpenAI, Anthropic, Google AI Studio and the rest, see API-key providers.
Checked against the gateway on 31 October 2026. Something wrong or missing? Tell us.