/v1/account/* use a provisioning key.
If you only need to call models, generate embeddings, create images, or use the chat-compatible APIs, your normal API key is enough.
Which key do I use?
| Request type | Use this key |
|---|---|
| Most API requests to models and generation endpoints | Standard API key |
Administrative requests under /v1/account/* | Provisioning key |
Standard API requests
Endpoints likeResponses, Embeddings, Images, Audio, Moderations, Chat Completions, and Messages accept your standard API key.
You can send it in either header:
Authorization: Bearer is the best default unless you already rely on x-api-key.
Account endpoints
Endpoints under/v1/account/* require a provisioning key instead of a regular inference key.
Use it in the bearer header:
- checking account balance
- retrieving account activity
- creating and managing API keys
Public or optional-auth endpoints
GET /v1/modelscan be called anonymously or with standard API auth.GET /v1/startupsis public metadata and does not require auth.
Standard request example
x-api-key, the same request works with:
Account request example
For provisioning-key-protected account endpoints, use the same bearer pattern:Common Mistakes
| Problem | Likely cause | Fix |
|---|---|---|
401 Unauthorized on inference endpoints | Missing or invalid API key | Recreate or resend a standard API key |
401 Unauthorized on /v1/account/* | Used a normal API key instead of a provisioning key | Use the provisioning key from the dashboard |
| Request works in one SDK but not another | Mixed base_url or header configuration | Ensure requests go to https://api.naga.ac/v1 |
| Browser-side key leakage risk | API key embedded in client-side code | Move calls behind your server or trusted backend |