# Vercel AI SDK

An MCP client with the streamable HTTP transport hands the tools to any model.

API key

## Setup

1.  1
    
    Create the client once and pass its tools to generateText or streamText.
    
    ```ts
    import { experimental_createMCPClient as createMCPClient, generateText } from "ai";
    import { StreamableHTTPClientTransport } from "@modelcontextprotocol/sdk/client/streamableHttp.js";
    
    const mcp = await createMCPClient({
      transport: new StreamableHTTPClientTransport(new URL("https://api.zomler.com/mcp"), {
        requestInit: { headers: { Authorization: `Bearer ${process.env.ZOMLER_API_KEY}` } },
      }),
    });
    
    const { text } = await generateText({
      model: "anthropic/claude-fable-5-1",
      tools: await mcp.tools(),
      prompt: "Compare @example and @sample on Instagram.",
    });
    await mcp.close();
    ```
    

## What you get

One tool per platform, plus `capabilities` and `check_balance`, at `https://api.zomler.com/mcp`. Every answer is [the envelope](https://zomler.com/docs/api/responses): what was measured, when, what was left out and why, and what it cost. Fresh fetches spend [credits](https://zomler.com/docs/api/credits); anything already held is free.

Keys are created under [API keys](https://app.zomler.com/api-keys) in the dashboard, where each one's calls and spend are listed. Revoking a key stops this client on its next call. The same key also calls the [REST API](https://zomler.com/docs/api).

[All clients](https://zomler.com/docs/mcp)

---

Source: https://zomler.com/docs/mcp/vercel-ai