Use wasmagent-js kernels with Vercel AI SDK

The wasmagent-js code-execution kernels (@wasmagent/kernel-quickjs, kernel-pyodide, kernel-wasmtime, kernel-remote) do not require the rest of wasmagent-js. You can drop them into any agent framework that lets you register a custom tool — including Vercel AI SDK.

This page shows how to expose QuickJSKernel as a Vercel AI SDK tool, giving you edge-safe sandboxed code execution that the AI SDK does not ship today.

Why bother

Vercel AI SDK 6 has world-class React DX, streaming UI primitives, and a default place in Next.js templates. What it does not have is a way to run model-generated code inside a real sandbox on the edge. node:vm is forbidden on Cloudflare Workers and Vercel Edge; OS-level sandboxes need a server.

@wasmagent/kernel-quickjs runs JavaScript inside QuickJS-in-WASM — language-level isolation, no node:vm, ~2 MB cold start. Perfect fill for that gap.

Install

npm install ai @ai-sdk/openai @wasmagent/kernel-quickjs quickjs-emscripten @jitl/quickjs-wasmfile-release-sync zod

Wire the kernel as a tool

import { generateText, tool } from "ai";
import { openai } from "@ai-sdk/openai";
import { QuickJSKernel } from "@wasmagent/kernel-quickjs";
import { z } from "zod";

const kernel = new QuickJSKernel();

const runJs = tool({
  description:
    "Run a JavaScript expression inside a sandbox. Returns the value of the last expression.",
  parameters: z.object({
    code: z.string().describe("JavaScript code; the value of the final expression is returned."),
  }),
  execute: async ({ code }) => {
    const result = await kernel.run(code);
    return { output: result.output, logs: result.logs };
  },
});

const { text } = await generateText({
  model: openai("gpt-4o"),
  tools: { runJs },
  prompt: "Compute the 12th Fibonacci number using the runJs tool.",
});

console.log(text);

That is the whole integration. The kernel runs anywhere Vercel AI SDK runs — Workers, Edge, Node — because QuickJS is just WASM.

What you get for free

• No node:vm requirement — works on Cloudflare Workers, Vercel Edge, Deno Deploy.
• Tight isolation — QuickJS is a separate VM; sandboxed code cannot touch your worker's globals or environment.
• Capability-controlled — pass a CapabilityManifest as the second argument to kernel.run(code, capabilities) to grant or revoke specific host imports.
• Drop-in tier upgrade — swap to @wasmagent/kernel-pyodide for Python or @wasmagent/kernel-remote for E2B microVMs without changing the tool wrapper.

See also

• Kernel decision tree — pick the right tier
• @wasmagent/kernel-quickjs README — package-level docs
• Use kernels with Mastra — the same idea for the Mastra framework