Agent Keys

The same models, from your own code

An Agent Key lets an agent, script, or app call Cognia models over an OpenAI-compatible API. It spends your balance, inside the limits you set for that key.

Shown once: The secret appears a single time when you create the key. Only a keyed hash is stored, so nobody can read it back, including us.
Spend limit: A cap in US dollars per day, per month, or for the key's lifetime. The cap is checked and reserved atomically, so parallel requests cannot go over it.
Request limit: A cap on how many requests the key may send to a model in the same period. Requests that are refused or released without charge do not count.
Rate and concurrency: Requests per minute and simultaneous requests are limited per key.
Model allowlist: Restrict a key to the models an agent actually needs.
Usage log: Every request made with a key: time, model, tokens, cost, and outcome.
Revoke: Revoking blocks new requests immediately. A request already running finishes and is settled normally.

curl https://usecognia.xyz/v1/chat/completions \
  -H "Authorization: Bearer $COGNIA_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "<model-id>",
    "messages": [{"role": "user", "content": "Hello"}]
  }'

import OpenAI from "openai";

const client = new OpenAI({
  apiKey: process.env.COGNIA_API_KEY,
  baseURL: "https://usecognia.xyz/v1",
});

Create an Agent Key API documentation