Last reviewed July 2026
Browser automation is essential for web scraping, testing, and workflow automation. While Claude Code doesn't have built-in browser control, MCP servers provide powerful browser automation capabilities that let you interact with web pages, take screenshots, fill forms, and more. Here are the best options available.
Live demo · no signup Let Claude drive a live web session
Not a headless browser. A real live web session, running in this page. Claude connects to it over MCP, then clicks and types on the page while you watch - and you can take over at any time. Free for 10 minutes, no account.
Recommended MCP Servers
Playwright MCP
microsoft/playwright-mcp
Official Microsoft Playwright MCP server, enabling LLMs to interact with web pages through structured accessibility snapshots.
Key Features
- Official Microsoft support and maintenance
- Structured accessibility snapshots for reliable element targeting
- Cross-browser support (Chromium, Firefox, WebKit)
- Headless and headed mode support
Installation
npx @anthropic-ai/playwright-mcpPlaywright MCP is the gold standard for browser automation with Claude. It's officially maintained by Microsoft, provides structured accessibility data that helps Claude understand page content, and supports all major browsers. The reliability and developer experience are unmatched.
Browser MCP (ByteDance)
agent-infra/mcp-server-browser
Browser automation capabilities using Puppeteer, supporting both local and remote browser connections.
Key Features
- Built on Puppeteer for Chrome/Chromium
- Remote browser connection support
- Local browser automation
- Part of UI-TARS desktop framework
Installation
npm install @agent-infra/mcp-server-browserByteDance's Browser MCP is a solid choice if you're working primarily with Chrome. It leverages Puppeteer's mature ecosystem and adds the ability to connect to remote browser instances, which is useful for testing across different environments.
BrowserMCP
browsermcp/mcp
Automate your local Chrome browser with simple commands.
Key Features
- Simple setup and configuration
- Works with your existing Chrome profile
- Lightweight and focused
- Good for quick automation tasks
Installation
npx browsermcpBrowserMCP is perfect for developers who want quick browser automation without complex setup. It works with your existing Chrome installation and profile, making it ideal for automating tasks you'd normally do manually in the browser.
Playwright MCP Server
executeautomation/mcp-playwright
An MCP server using Playwright for browser automation and web scraping.
Key Features
- Full Playwright feature set
- Web scraping optimized
- Screenshot and PDF generation
- Form automation support
Installation
npx @executeautomation/mcp-playwrightThis community-maintained Playwright server is tailored for web scraping use cases. It provides convenient methods for extracting data from web pages and handles common scraping patterns well. A good choice if your primary goal is data extraction.
Quick Comparison
| MCP | Stars | Best For | Pros | Cons |
|---|---|---|---|---|
| Playwright MCP | 24k+ | Production automation | Official Microsoft support | Larger footprint |
| Browser MCP (ByteDance) | 19k+ | Chrome automation | Remote browser support | Chrome only |
| BrowserMCP | 5k+ | Quick local tasks | Simplest setup | Limited features |
| mcp-playwright | 5k+ | Web scraping | Scraping optimized | Community maintained |
Getting Started
To get started with browser automation in Claude Code:
Choose your MCP: For most users, Playwright MCP is recommended for its reliability and official support.
Install the server:
npx @anthropic-ai/playwright-mcpAdd to Claude Code config: Add the MCP server to your claude_desktop_config.json
Start automating: Ask Claude to navigate to pages, click buttons, fill forms, or extract data.
Example prompt: "Go to github.com and take a screenshot of the trending repositories page."
Related Guides
Frequently Asked Questions
Can Claude control a real browser that a human can take over?
Yes. Most MCP browser servers drive a fresh, headless browser the user never sees. Webfuse instead connects your agent to a real, live browser session over MCP, with no install or extension, and a human can take over at any moment. That suits logged-in or high-stakes tasks where you want a person able to step in.
What is the difference between headless browser automation and a live session?
Headless automation launches a throwaway browser that runs unattended, which is efficient for scraping and tests. A live session runs in a real, shareable browser where the agent and a human can act on the same page. Live sessions fit interactive, authenticated, or supervised workflows; headless fits bulk, unattended jobs.
