Mastra

Use the object model config for a custom endpoint:

// @mastra/core
import { Agent } from "@mastra/core/agent";

export const nitoAgent = new Agent({
  name: "assistant",
  instructions: "Answer concisely.",
  model: {
    url: "https://app.trynito.ai/v1",   // base URL, WITH /v1 (not the /chat/completions path)
    id: "nito/moonshotai/kimi-k3",                         // provider/model form: a label of your choice, then the catalog ID
    apiKey: process.env.NITO_API_KEY,            // your Nito key, "z_sk_..."
  },
});

A bare string such as model: "openai/..." carries no url, so the router resolves it to the built-in provider for that prefix. Use the object form for Nito.

Or an AI SDK provider instance, which Mastra accepts anywhere a model is expected:

import { createOpenAI } from "@ai-sdk/openai";
const nito = createOpenAI({ baseURL: "https://app.trynito.ai/v1", apiKey: process.env.NITO_API_KEY });
// model: nito.chat("moonshotai/kimi-k3")

id: "nito/z-ai/glm-5.2:confidential",

Privacy

Use a catalog ID, not an OpenAI name. For attested hardware use a :confidential variant; GET /v1/models lists their :confidential IDs. Encrypted needs client-side encryption and extra headers, which these frameworks do not add. See Privacy Levels.

Framework Integrations

The shared pattern.

OpenAI Compatibility

What differs.

Privacy Levels

The four levels.

Basics

Features

Server-Side Tools

Multimodal

Best Practices

SDKs

API Reference

Privacy

Framework Integrations

OpenAI Compatibility

Privacy Levels

​Privacy

​Related

Framework Integrations

OpenAI Compatibility

Privacy Levels

Privacy

Related