Use this file to discover all available pages before exploring further.
There’s a new way to automate the web: by prompting.But a single prompt fails on complicated scripts. Stagehand lets you prompt at multiple levels of granularity—from agents to single actions—and interweave Playwright or traditional frameworks when you need them.Use any major model provider. 22k GitHub stars, 700k+ weekly downloads. Get started in 5 minutes.
This quickstart uses Model Gateway — LLM requests for Stagehand are routed through your Browserbase API key, so you don’t need a separate model provider account. You can also bring your own API key if you prefer.
pnpm add @browserbasehq/stagehandnpm pkg set type=module
import { Stagehand } from "@browserbasehq/stagehand";const stagehand = new Stagehand({ env: "BROWSERBASE", model: "google/gemini-3-flash-preview",});await stagehand.init();const page = stagehand.context.pages()[0];await page.goto("https://news.ycombinator.com");// Let AI clickawait stagehand.act("click on the comments link for the top story");// Extract structured dataconst data = await stagehand.extract("extract the title and points of the top story");console.log(data);await stagehand.close();
Save as script.ts, then run:
pnpm dlx tsx script.ts
uv inituv add stagehand
from stagehand import Stagehanddef main(): client = Stagehand() response = client.sessions.start(model_name="google/gemini-3-flash-preview") session_id = response.data.session_id print(f"Session started: {session_id}") client.sessions.navigate(id=session_id, url="https://news.ycombinator.com") print("Navigated to Hacker News") # Let AI click act_result = client.sessions.act(id=session_id, input="click on the comments link for the top story") print(f"Act result: {act_result}") # Extract structured data result = client.sessions.extract( id=session_id, instruction="extract the title and points of the top story" ) print(f"Extracted: {result}") client.sessions.end(id=session_id)if __name__ == "__main__": main()
