Skip to main content
Corvex uses a single customer key type: a virtual key with the sk-corvex- prefix. Send it as an HTTP Bearer token on the OpenAI-compatible endpoints, or as an x-api-key header on the Anthropic-compatible /v1/messages surface.

Virtual keys (sk-corvex-*)

Use your virtual key on every inference endpoint: /v1/chat/completions, /v1/completions, /v1/responses, /v1/messages, /v1/messages/count_tokens, and /v1/models, plus the usage endpoint /api/v1/usage. (A /v1/embeddings endpoint also exists, but no embedding model is currently deployed.)
Use the Authorization: Bearer header on the OpenAI-compatible endpoints. On the Anthropic-compatible /v1/messages and /v1/messages/count_tokens endpoints, send your key as an x-api-key header (as the Anthropic SDK and Claude Code expect). The current prefix is sk-corvex-. Legacy sk-bf-* keys minted before the cutover are still accepted, so existing integrations keep working — but new keys are issued with the sk-corvex- prefix. Your key is scoped to your account, and the models it can reach and any usage limits are managed for you from the dashboard.
Treat your virtual key like a password. Keep it server-side, never commit it to source control, and don’t ship it to client devices or browsers. If a key is exposed, revoke it from the dashboard and mint a new one.

Where to mint keys

Mint virtual keys from the Corvex dashboard — sign in at tokenfactory.corvex.cloud and create a key from your API-keys section. The key value (sk-corvex-...) is shown once — copy it immediately and store it somewhere safe.
Keys are created and revoked from the dashboard. There is no API to create or manage keys programmatically.

Rate limits

If you exceed your rate limit you’ll receive an HTTP 429 Too Many Requests response with a Retry-After header indicating how long to wait before retrying. Back off for the indicated interval, then retry your request. You can review your current usage from the dashboard, or fetch it programmatically from /api/v1/usage using the same virtual key.

Going further

API Reference

Full endpoint list and request/response schemas.

Errors

Full HTTP status code list.