Start with the same SDK shape your tools already understand, then add policies and quotas behind the gateway.
import OpenAI from "openai"; const client = new OpenAI({ apiKey: process.env.COVIBE_KEY, baseURL: "https://covibe.me/v1" }); await client.chat.completions.create({ model: "coding-prod", messages: [{ role: "user", content: "Fix this test" }] });
{
"coding-prod": "cost-aware + fallback",
"coding-fast": "low-latency default",
"demo-safe": "low budget, no premium"
}{
"model": "coding-prod",
"policy": "cost-aware",
"budget": "team-agents",
"fallbacks_available": 2,
"quota_remaining": "68%"
}Run self-hosted with Docker, keep raw provider ports private, expose client API through HTTPS only.