Mastra

Mastra agents take a Vercel AI SDK model. ai-lcr returns one — so a Mastra agent gets cheapest-first routing and fallback by changing a single field.

Mastra is a TypeScript agent framework built on the Vercel AI SDK. A Mastra agent's model field is a standard AI SDK LanguageModel — exactly what lcr("...") returns. So an agent gets least-cost routing and fallback with one field changed.

import { Agent } from "@mastra/core/agent";
import { createLCR } from "ai-lcr";
import { createOpenAICompatible } from "@ai-sdk/openai-compatible";
import { openai } from "@ai-sdk/openai";

const openrouter = createOpenAICompatible({
  name: "openrouter",
  baseURL: "https://openrouter.ai/api/v1",
  apiKey: process.env.OPENROUTER_API_KEY,
});

const lcr = createLCR({
  autoSort: true,
  models: {
    "support-llm": [
      { model: openrouter("openai/gpt-4o-mini"), label: "openrouter", cost: { input: 0.15, output: 0.6 } },
      { model: openai("gpt-4o-mini"), label: "openai-direct", cost: { input: 0.15, output: 0.6 } },
    ],
  },
  onCost: ({ provider, costUsd }) => console.log(`support agent · ${provider}: $${costUsd.toFixed(6)}`),
});

export const support = new Agent({
  name: "support",
  instructions: "You are a concise support agent.",
  model: lcr("support-llm"), // ← the only change
});

Tool calls, memory, workflows, and streaming inside Mastra are unaffected — they operate on the model interface, which ai-lcr preserves.

Dynamic model selection

Mastra also supports a function-form model. ai-lcr composes with it — return lcr("...") based on runtime context:

export const support = new Agent({
  name: "support",
  instructions: "...",
  model: ({ runtimeContext }) =>
    runtimeContext.get("tier") === "premium" ? lcr("smart") : lcr("cheap"),
});

Dynamic model selection

On this page