Does Playwright MCP work with Claude Code?

Yes. Any MCP-compatible client including Claude Code, Claude Desktop, and Cursor can use the Playwright MCP server.

MCP Servers

Playwright MCP Server — Browser Automation for AI

Learn how to set up and use the Playwright MCP server to give AI assistants full browser automation capabilities — navigate pages, fill forms, take screenshots, and extract data.

The Playwright MCP server gives AI assistants like Claude the ability to control a real web browser — navigating pages, clicking elements, filling forms, taking screenshots, and extracting structured data. It turns natural language into browser automation without writing test scripts.

What Is the Playwright MCP Server?

The Playwright MCP server wraps Microsoft's Playwright browser automation library in a Model Context Protocol interface. Instead of writing JavaScript test scripts to automate browsers, you describe what you want in plain English, and the AI assistant translates your intent into browser actions through MCP tool calls.

This is particularly powerful for tasks like web scraping, form filling, visual regression testing, site monitoring, and research workflows. The AI handles the browser while you focus on what to do with the results. Under the hood, Playwright supports Chromium, Firefox, and WebKit — so you get cross-browser compatibility out of the box.

language

Navigate & Interact

Click, type, scroll, and navigate across any web page

screenshot_monitor

Screenshots & PDFs

Capture full-page screenshots or export pages as PDF

data_object

Data Extraction

Extract text, tables, links, and structured data from pages

Installation and Setup

The Playwright MCP server is available as an npm package. Install it and configure it in one step:

npm install -g @anthropic/mcp-playwright

Then add it to your Claude Desktop configuration file:

{
  "mcpServers": {
    "playwright": {
      "command": "npx",
      "args": ["@anthropic/mcp-playwright"],
      "env": {
        "PLAYWRIGHT_HEADLESS": "true"
      }
    }
  }
}

Set PLAYWRIGHT_HEADLESS to false if you want to watch the browser work in real time — useful for debugging and demonstrations. In production or server environments, headless mode is faster and uses fewer resources.

Available Tools and Capabilities

Once connected, the Playwright MCP server exposes a rich set of browser automation tools to the AI assistant:

Tool	What It Does	Use Case
`navigate`	Opens a URL in the browser	Starting any web workflow
`click`	Clicks an element by selector or text	Clicking buttons, links, dropdowns
`fill`	Types text into form fields	Login forms, search bars, data entry
`screenshot`	Captures a screenshot of the current page	Visual verification, bug reports
`get_text`	Extracts all visible text from the page	Content scraping, research
`evaluate`	Runs JavaScript in the browser context	Complex DOM queries, custom extraction

Practical Use Cases

The Playwright MCP server is most valuable when you need the AI to interact with live web content rather than just processing static text. Here are the workflows where it shines:

monitoring

Site Monitoring and QA

Ask Claude to visit your site, check that key pages load correctly, verify form submissions work, and report any visual regressions — all through conversation.

travel_explore

Web Research and Data Collection

Navigate to competitor sites, extract pricing tables, compare product features, and compile the results — without manually visiting dozens of pages.

edit_note

Form Automation

Fill out repetitive web forms — CRM entries, support tickets, data imports — by describing the data and letting the AI handle the browser interaction.

bug_report

Bug Reproduction

Describe a bug report to Claude and let it navigate to reproduce the issue, take screenshots of the broken state, and document the steps — ideal for QA workflows.

Building a Custom Playwright MCP Server

If you need capabilities beyond the standard server, you can build a custom MCP server that wraps Playwright with your specific automation logic. Here is a Python example using FastMCP and Playwright's async API:

from mcp.server.fastmcp import FastMCP
from playwright.async_api import async_playwright

mcp = FastMCP("browser-tools")

@mcp.tool()
async def scrape_page(url: str, selector: str = "body") -> str:
    """Navigate to a URL and extract text from a CSS selector."""
    async with async_playwright() as p:
        browser = await p.chromium.launch(headless=True)
        page = await browser.new_page()
        await page.goto(url, wait_until="networkidle")
        content = await page.inner_text(selector)
        await browser.close()
        return content

@mcp.tool()
async def take_screenshot(url: str) -> str:
    """Take a full-page screenshot and return the file path."""
    async with async_playwright() as p:
        browser = await p.chromium.launch(headless=True)
        page = await browser.new_page()
        await page.goto(url, wait_until="networkidle")
        path = f"/tmp/screenshot_{hash(url)}.png"
        await page.screenshot(path=path, full_page=True)
        await browser.close()
        return f"Screenshot saved: {path}"

Install the dependencies with pip install mcp playwright && playwright install chromium, then connect it to Claude using your standard MCP configuration.

Security and Performance Tips

check_circleRun headless in production — saves memory and CPU; only use headed mode for debugging.

check_circleSet navigation timeouts — prevent the server from hanging on slow or unresponsive pages.

check_circleSandbox the browser — run in a container or restricted network environment to limit exposure.

check_circleReuse browser contexts — avoid launching a new browser for every request; use persistent contexts for session-based workflows.

check_circleBlock unnecessary resources — disable images, fonts, and analytics scripts when scraping text-only content.

Frequently Asked Questions

Can the Playwright MCP server handle JavaScript-heavy sites?

Yes. Playwright uses real browser engines (Chromium, Firefox, WebKit) that execute JavaScript fully. Single-page apps, dynamic content, and client-side rendering all work as expected.

Does it work with Claude Code or just Claude Desktop?

Both. Any MCP-compatible client can use the Playwright server. Claude Code, Claude Desktop, Cursor, and other MCP clients all support it through the same configuration.

Can I use Playwright MCP for web scraping at scale?

For small-to-medium scraping tasks, it works well. For high-volume scraping, you are better off writing dedicated scripts — the MCP overhead and AI inference costs make it expensive at scale. Playwright MCP is best for interactive, judgment-heavy tasks where the AI adds value beyond simple automation.

Last updated: 2026 • Browse all courses