Automate Web Interactions with Cloud Browsers
Control cloud browsers with AI: navigate, click, fill forms, take screenshots, and extract structured data via Browserbase and Stagehand.
Why it matters
Leverage cloud browser automation to enable LLMs to interact with web pages, extract structured data, and perform automated actions with precision.
Outcomes
What it gets done
Control cloud browsers via Browserbase for web automation.
Extract structured data from any webpage.
Navigate, click, and fill forms on web pages.
Take full-page and element screenshots.
Install
Add it to your toolbox
Run in your project directory:
curl -fsSL https://spark.entire.vc/get/vb-browserbase | bash Capabilities
Tools your agent gets
Navigate to a specified URL in the browser
Capture a full-page or element screenshot from the browser
Extract structured data from a webpage
Click on an element in the browser
Fill form fields with specified values
Create a new browser session
Close an active browser session
Manage and control browser sessions
Overview
Browserbase MCP
A cloud browser automation MCP server built on Browserbase and Stagehand, giving AI assistants navigation, form interaction, screenshots, and structured data extraction across real browsers. Reach for this when an AI agent needs to interact with a real, rendered browser session - not just static scraped HTML - to navigate, fill forms, or extract data.
What it does
The Browserbase MCP server provides cloud browser automation using Browserbase and Stagehand, enabling LLMs to interact with web pages, take screenshots, extract information, and perform automated actions with atomic precision. It supports browser automation (navigate, click, fill forms), structured data extraction from any webpage, full-page and element screenshots, and session management (create, manage, close browser sessions).
It is model-flexible (OpenAI, Claude, Gemini, and more) and includes vision support via annotated screenshots for complex DOMs. Stagehand v3 adds targeted extraction across iframes and shadow roots, CSS selector support with improved element targeting, multi-browser support (Playwright, Puppeteer, Patchright), and built-in primitives (page, locator, frameLocator, deepLocator), plus 20-40% faster performance via automatic caching.
npm install -g @browserbasehq/mcp-server-browserbase
When to use - and when NOT to
Use this when you want an AI assistant to control a real cloud browser - navigating pages, filling forms, clicking through flows, taking screenshots, and extracting structured data like product prices - rather than working from static scraped HTML.
Do NOT use this without a BROWSERBASE_API_KEY and BROWSERBASE_PROJECT_ID (plus a model API key such as GEMINI_API_KEY) - all are required for the local server configuration. Advanced stealth mode requires the Scale Plan.
Inputs and outputs
Inputs are navigation targets, DOM selectors, form data, or extraction instructions; outputs are page screenshots, extracted structured data, and session state.
Integrations
Browserbase (cloud browser infrastructure) and Stagehand (browser automation framework), supporting Playwright, Puppeteer, and Patchright under the hood. Configurable via flags including --proxies, --advancedStealth, --keepAlive, --contextId, --browserWidth/--browserHeight, and --modelName (default gemini-2.0-flash).
Who it's for
Developers building AI agents that need real browser interaction - form-filling, navigation, screenshots, and structured data extraction - rather than static web scraping.
Source README
The Browserbase MCP server provides cloud browser automation capabilities using Browserbase and Stagehand. It enables LLMs to interact with web pages, take screenshots, extract information, and perform automated actions with atomic precision.
Installation
npm install -g @browserbasehq/mcp-server-browserbase
Configuration
Remote Server (Recommended)
Get a hosted URL from Smithery with LLM costs included:
{
"mcpServers": {
"browserbase": {
"type": "http",
"url": "your-smithery-url.com"
}
}
}
Local Server
{
"mcpServers": {
"browserbase": {
"command": "npx",
"args": ["@browserbasehq/mcp-server-browserbase"],
"env": {
"BROWSERBASE_API_KEY": "your-api-key",
"BROWSERBASE_PROJECT_ID": "your-project-id",
"GEMINI_API_KEY": "your-gemini-key"
}
}
}
}
Features
| Feature | Description |
|---|---|
| Browser Automation | Control cloud browsers via Browserbase |
| Data Extraction | Extract structured data from any webpage |
| Web Interaction | Navigate, click, and fill forms |
| Screenshots | Full-page and element screenshots |
| Model Flexibility | Supports OpenAI, Claude, Gemini, and more |
| Vision Support | Annotated screenshots for complex DOMs |
| Session Management | Create, manage, and close browser sessions |
| High Performance | 20-40% faster with automatic caching (v3) |
Stagehand v3 Features
- Targeted extraction across iframes and shadow roots
- CSS selector support with improved element targeting
- Multi-browser support (Playwright, Puppeteer, Patchright)
- Built-in primitives:
page,locator,frameLocator,deepLocator - Experimental features with
--experimentalflag
Configuration Options
| Flag | Description |
|---|---|
--proxies |
Enable Browserbase proxies |
--advancedStealth |
Enable advanced stealth mode (Scale Plan) |
--keepAlive |
Enable keep-alive sessions |
--contextId |
Specify a Browserbase Context ID |
--browserWidth |
Viewport width (default: 1024) |
--browserHeight |
Viewport height (default: 768) |
--modelName |
AI model (default: gemini-2.0-flash) |
--experimental |
Enable experimental features |
Usage Example
Claude, navigate to example.com, take a screenshot,
and extract all product prices from the page.
Resources
FAQ
Common questions
Discussion
Questions & comments · 0
Sign In Sign in to leave a comment.