Authentication

plm_ API keys, dashboard JWT, public unauthenticated routes, plmk_ provisioning, per-protocol headers.

Choose a credential by task

plm_: model inference requests
JWT: dashboard key and account management
plmk_: child-key provisioning for automation
Public GET routes: no authentication

plm_#

Inference uses Authorization: Bearer $PLEUM_API_KEY. The secret is shown once; only a hash is stored. Keys

Authorization

Authorization: Bearer plm_xxxxxxxxxxxxxxxx

JWT#

Dashboard session token. Used for key CRUD, usage, routing policies, and other account admin.

Public#

GET /v1/models, GET /v1/fx-rate, GET /v1/stats/* — No authentication needed. Catalog, FX, and public stats.

plmk_#

Provisioning keys mint plm_ inference keys. Do not send them as inference Bearer. Provisioning

Per-protocol headers#

OpenAI SDK: apiKey + baseURL=…/v1. Anthropic SDK: root origin plus x-api-key or Bearer. Gemini SDK: root origin plus api_key. Compatibility

Troubleshoot 401 and 403#

For 401, check the credential for the operation (a plm_ inference key or JWT management token), the Authorization: Bearer format, and the API base URL. For 403, check the key's scopes or the account permission. See error codes · key scopes.

GET /v1/credits accepts JWT or an API key. Credits