Setup
Three lines change in your app.
You change the address your app sends AI requests to, the key and the model name. The rest of your code stays the same.
app.ts ยท three lines change
const client = new OpenAI({ baseURL: "https://api.knectar.dev/openai/v1", // 1 knectar address apiKey: process.env.OPENAI_API_KEY, // 2 your knectar key});client.chat.completions.create({ model: "openai/gpt-4o-mini", // 3 provider/model messages,});Your codeAddress ends inKey goes inNotes
OpenAI SDK /openai/v1 apiKey The SDK adds its own paths. Write the model as provider/model.
Vercel AI SDK, LangChain or plain fetch /v1 Authorization: Bearer Anything that uses the OpenAI format and lets you set the address.
Anthropic SDK /anthropic apiKey Sent as x-api-key.
Status check /health no key Tells you knectar is up. It does not call a model.
Model names
Every model name starts with its AI provider, like openai/gpt-4o-mini. Backup models are written the same way.
Without an SDK, send your knectar key in Authorization: Bearer, x-api-key, x-goog-api-key, api-key or x-bf-vk.
- openai/gpt-4o-mini
- anthropic/claude-haiku-4-5
- gemini/gemini-2.5-flash
- groq/llama-3.3-70b-versatile
- openrouter/<vendor>/<model>
Connect your own key for OpenAI, Anthropic, Gemini, Groq or OpenRouter. One OpenRouter key reaches hundreds of models. knectar does not sell AI access, so each AI provider bills you as it does now.