Browser

Browser automation with AgentBrowser and Stagehand.

Topics

Overview - Introduction to browser automation
AgentBrowser - Browser control for agents
Stagehand - Visual page understanding

Browser Overview

Browser automation enables agents to interact with web pages and web-based applications.

Capabilities

Navigate to URLs
Fill forms and click buttons
Extract data from pages
Take screenshots
Execute JavaScript
Handle multi-tab workflows

Browser Providers

AgentBrowser - Simple browser automation
Stagehand - AI-powered visual understanding
Playwright - Low-level browser control

Getting Started

const browser = await AgentBrowser.launch({
  headless: true,
  viewport: { width: 1920, height: 1080 },
});

const page = await browser.newPage();
await page.goto('https://example.com');

AgentBrowser

Browser automation for agents.

Creating a Browser Agent

import { AgentBrowser } from '@mastra/core';

const browserAgent = new AgentBrowser({
  name: 'web-researcher',
  instructions: 'You browse the web to find information',
  browser: {
    headless: true,
  },
});

Basic Operations

// Navigate to a page
await browserAgent.goto('https://example.com');

// Click an element
await browserAgent.click('button.submit');

// Fill a form
await browserAgent.fill('input[name="email"]', 'user@example.com');

// Extract content
const title = await browserAgent.text('h1');
const links = await browserAgent.links('a');

Agent Integration

const agent = mastra.getAgent('myAgent');

browserAgent.useAgent(agent);

await browserAgent.start();

Screenshot

const screenshot = await browserAgent.screenshot();
console.log(screenshot.url);

Stagehand

AI-powered browser automation with visual understanding.

Overview

Stagehand uses AI to understand page structure and elements, making it easier to automate complex web pages.

Getting Started

import { StagehandBrowser } from '@mastra/stagehand';

const browser = await StagehandBrowser.launch({
  headless: true,
  model: 'gpt-4-vision',
});

Natural Language Commands

const page = await browser.newPage();
await page.goto('https://example.com');

// Use natural language to interact
await page.act('Click the "Sign Up" button');
await page.act('Fill in the registration form with email and password');
await page.act('Submit the form');

Visual Understanding

// Stagehand understands page visually
const elements = await page.findElements({
  type: 'button',
  visible: true,
  containsText: 'Submit',
});

Extract Data

const data = await page.extract({
  title: 'h1.product-title',
  price: 'span.price',
  description: 'div.description',
});

Limitations

Stagehand currently has beta support. For production use, consider AgentBrowser or Playwright.