Pull structured data from any web page with plain English

What you get

A Node script that opens a real Chrome window on your machine, goes to a web page, and uses plain-English instructions to pull typed data off it (extract), click something (act), and list what can be clicked (observe). No cloud browser account needed, only a model key.

Prerequisites

Node.js 22.18 or newer and Google Chrome installed on your machine (Stagehand drives it through the Chrome DevTools Protocol).
An OpenAI API key in OPENAI_API_KEY.

Steps

Create a project and install the SDK. Keep Zod on 4.4.x; the docs say newer Zod minors can cause TypeScript errors with extract().

Shell

mkdir my-stagehand-app && cd my-stagehand-app
pnpm init -y
pnpm install @browserbasehq/stagehand@4.2.1 zod@4.4.3

Create index.ts:

typescript

import { localBrowser, Stagehand } from "@browserbasehq/stagehand";
import { z } from "zod/v4";

async function main() {
  const browser = await localBrowser.launch();
  try {
    const stagehand = await Stagehand.create({
      browser,
      model: {
        modelName: "openai/gpt-5.6-sol",
        apiKey: process.env.OPENAI_API_KEY,
      },
    });
    console.log("Stagehand session started");
    try {
      const [page] = await browser.context.pages();

      await page.goto("https://stagehand.dev");

      const extractResult = await stagehand.extract(
        "Extract the value proposition from the page.",
        z.object({ valueProposition: z.string() }),
      );
      console.log("Extract result:\n", extractResult.data);

      await stagehand.act("Click the 'Evals' button.");

      const observeResult = await stagehand.observe("What can I click on this page?");
      console.log("Observe result:\n", observeResult.data);
    } finally {
      await stagehand.close();
    }
  } finally {
    await browser.close();
  }
}

main().catch((err) => {
  console.error(err);
  process.exit(1);
});

Stagehand never reads environment variables for you. The script reads OPENAI_API_KEY itself and passes it in. A local browser cannot use Browserbase's Model Gateway, so you must give it a model and key.

Test it

Shell

export OPENAI_API_KEY="sk-..." # Your model provider API key
pnpm dlx tsx index.ts          # Run the example script

A Chrome window opens, visits stagehand.dev, and the terminal prints the extracted valueProposition, then the observe result after it clicks Evals.

Not confirmed

The docs do not say how much a run costs in model tokens.
The quickstart uses OpenAI; other providers (ANTHROPIC_API_KEY, GOOGLE_GENERATIVE_AI_API_KEY) are listed on the models page but not shown in this exact script.
Cloud mode (browserbase.launch() with BROWSERBASE_API_KEY) is mentioned but not covered here.
The docs run the script with pnpm dlx tsx and do not pin a tsx version.
Zod is pinned to 4.4.3 (the docs ask for 4.4.x) rather than the newest Zod release.

What you'll need

What you get

Prerequisites

Steps

Test it

Not confirmed