TypeScript AI toolkit
How to use Vercel AI SDK with Rapid Router
The AI SDK's OpenAI provider accepts a baseURL. Create one provider instance for Rapid Router and pass model strings from the catalog to streamText, generateText and the React hooks.
Setup
- Install ai and @ai-sdk/openai.
- Create the provider with createOpenAI as below.
- Pass rapidrouter("<model string>") wherever the SDK wants a model.
import { createOpenAI } from "@ai-sdk/openai";
import { streamText } from "ai";
const rapidrouter = createOpenAI({
baseURL: "https://rapidrouter.ai/api/v1",
apiKey: process.env.RAPIDROUTER_API_KEY,
});
const result = streamText({
model: rapidrouter("anthropic/claude-sonnet-5"),
prompt: "Write a haiku about the monsoon",
});Field names as in Vercel AI SDK at the time of writing; their UI moves, the three values do not. Vercel AI SDK site.
Pick a model
The catalog lists every live model with its price per million tokens. For a coding tool, a strong model for edits and a cheap one for questions is the usual pair; the cost calculator shows what a month of your usage costs on each.
Common questions
Which model string do I put in Vercel AI SDK?
Any id from the catalog, in the vendor/model shape, for example anthropic/claude-sonnet-5. The models page lists every live id with its price; copy it exactly.
Does Vercel AI SDK need anything else to work with Rapid Router?
No. Rapid Router exposes the OpenAI chat completions API at https://rapidrouter.ai/api/v1, which is the shape Vercel AI SDK already sends. Streaming, tool calling and JSON mode work where the model supports them.
How is usage from Vercel AI SDK billed?
Per request at the provider's list price per million tokens, debited from your prepaid balance. Create a separate key for Vercel AI SDK and the dashboard shows its spend on its own; you can cap it monthly.
Is there a free allowance to test with?
Yes. Every account gets a daily allowance of free requests on eligible models before any top-up, so the first session costs nothing.