Linked Playwright Mcp Server

MCP server providing full browser automation capabilities through the Playwright framework by Automata Labs.

When to Use This MCP Server

Connect this server when you need to:

• Automate cross-browser testing across Chromium, Firefox, and WebKit from within your Claude Code workflow
• Perform complex multi-step browser interactions including file uploads, downloads, and iframe navigation
• Generate test scripts or perform end-to-end testing with Playwright's powerful selector engine and auto-wait mechanisms

Consider alternatives when:

• You need cloud-hosted browser sessions and prefer not to manage local browser installations
• Simple HTTP requests or lightweight scraping would suffice for your data extraction needs

Quick Start

Configuration

Copy
name: linked-playwright-mcp-server
type: mcp
category: browser_automation

Example Connection

Copy
claude mcp add linked-playwright-mcp-server -- npx -y @automatalabs/mcp-server-playwright

Available Tools

navigate:        Navigate the browser to a specified URL
click:           Click on elements using Playwright selectors
fill:            Fill form inputs with specified text values
screenshot:      Capture full-page or element-level screenshots
evaluate:        Execute JavaScript code in the browser context
wait-for:        Wait for specific elements or conditions before proceeding

Core Concepts

Playwright MCP Overview

Aspect	Details
Protocol	MCP over stdio transport
Browser Support	Chromium, Firefox, and WebKit engines
Selector Engine	CSS, XPath, text, and role-based selectors
Auto-Wait	Built-in actionability checks before interactions
Network Control	Request interception, mocking, and routing
Installation	Single npx command, auto-downloads browsers

Playwright Integration Architecture

┌─────────────┐     ┌──────────────────┐     ┌────────────────┐
│ Claude Code  │────▶│  Playwright MCP  │────▶│  Chromium /    │
│  (Client)    │◀────│  Server          │◀────│  Firefox /     │
└─────────────┘     └──────────────────┘     │  WebKit        │
                           │                  └────────────────┘
                    ┌──────┴──────┐
                    │  Playwright │
                    │  Core API   │
                    └─────────────┘

Configuration

Parameter	Type	Default	Description
browser	string	chromium	Browser engine to use: chromium, firefox, or webkit
headless	boolean	true	Run the browser without a visible window
viewport_width	integer	1280	Width of the browser viewport in pixels
viewport_height	integer	720	Height of the browser viewport in pixels
slow_mo	integer	0	Milliseconds to slow down each operation for debugging

Best Practices

1. Use Auto-Wait Instead of Manual Delays - Playwright automatically waits for elements to be actionable before performing clicks or fills. Avoid adding manual sleep calls, which make tests flaky and slow.

2. Prefer Role-Based Selectors - Use role selectors like getByRole('button', { name: 'Submit' }) for resilient test scripts. These survive UI redesigns better than CSS class-based selectors that change frequently.

3. Run Multiple Browsers for Compatibility - Test across Chromium, Firefox, and WebKit to catch cross-browser issues early. The MCP server supports switching engines via configuration without code changes.

4. Intercept Network Requests for Testing - Use Playwright's network interception to mock API responses during testing. This isolates your browser tests from backend availability and produces deterministic results.

5. Capture Traces for Failed Interactions - Enable Playwright tracing to record a detailed timeline of actions, network requests, and DOM snapshots. These traces are invaluable for debugging automation failures after the fact.

Common Issues

1. Browser Binary Not Found - On first run, Playwright needs to download browser binaries. Run npx playwright install to pre-download browsers before connecting the MCP server.

2. Element Not Visible - Playwright refuses to click hidden elements by default. Check for overlaying modals, cookie banners, or CSS display: none rules that may be obscuring the target element.

3. Timeout Waiting for Selector - The default timeout may be too short for slow-loading pages. Increase the timeout in your tool call parameters or ensure the page has finished loading before interacting.