Skip to content

Overview

OpenAI-compatible inference, paid per reply with account credits or USDC.

Katara serves chat completions from Katara’s own models and, for requests paid in USDC, from independent GPU providers in Katara Community. The API follows the OpenAI chat completions shape, so existing clients work by changing the base URL and the key.

Authentication is a bearer key created in the portal. The key’s environment decides where a request is served and paid: a Live key on the Live network, in USDC or from Account Credits, and a Test key on the Test network, in test USDC. Networks names the chain and contracts of each. The settled price comes back on the response.

Format validation errors use the endpoint’s documented envelope. Payment challenges and admission refusals retain their status, body and payment headers. The code field is stable; the message is for people. Every response carries an x-katara-request-id header. Quote it when you contact support.

Authorization: Bearer kat_live_… or Bearer kat_test_…. A key made in a Live environment starts kat_live_, one made in a Test environment kat_test_; the environment, not the prefix, decides the network, so a key issued before prefixes followed the environment keeps working where it did. Create keys in the portal under API Keys with an organization, an environment, the cortex:inference scope and an expiry. A key is shown once. Subsequent requests require current, unrevoked authority.

Security scheme type: http