Linked Browsermcp

Lightweight MCP server for browser automation that lets AI applications navigate the web, fill out forms, and interact with page elements.

When to Use This MCP Server

Connect this server when you need to:

• Quickly automate browser tasks like form submissions, clicking buttons, and navigating between pages without heavy setup
• Perform web-based testing or verification steps as part of a Claude Code development workflow
• Extract visual or rendered content from pages that require JavaScript execution to display properly

Consider alternatives when:

• You need enterprise-grade browser automation with session recording and parallel execution capabilities
• Your use case only involves fetching raw HTML or calling REST APIs without browser rendering

Quick Start

Configuration

Copy
name: linked-browsermcp
type: mcp
category: browser_automation

Example Connection

Copy
claude mcp add linked-browsermcp -- npx @browsermcp/mcp@latest

Available Tools

navigate:      Load a specified URL in the browser
click:         Click on a page element by CSS selector or description
type:          Enter text into form fields and input elements
screenshot:    Capture the current viewport as an image
get-text:      Extract text content from the visible page
evaluate:      Run arbitrary JavaScript in the page context

Core Concepts

BrowserMCP Overview

Aspect	Details
Protocol	MCP over stdio transport
Installation	Zero-config via npx, no local dependencies required
Browser Engine	System Chromium or bundled browser instance
State Persistence	Session state maintained across sequential tool calls
Authentication	No API keys required for the MCP server itself
Use Cases	Form testing, page verification, content extraction

Integration Architecture

┌─────────────┐     ┌──────────────────┐     ┌──────────────┐
│ Claude Code  │────▶│   BrowserMCP     │────▶│   Chromium   │
│  (Client)    │◀────│   Server (npx)   │◀────│   Instance   │
└─────────────┘     └──────────────────┘     └──────────────┘
                                                    │
                                             ┌──────┴──────┐
                                             │   Web Page   │
                                             │   (Target)   │
                                             └─────────────┘

Configuration

Parameter	Type	Default	Description
headless	boolean	true	Run browser in headless mode without a visible window
timeout	integer	30000	Maximum wait time in milliseconds for page navigation
viewport_width	integer	1280	Browser viewport width in pixels
viewport_height	integer	720	Browser viewport height in pixels
user_agent	string	auto	Custom User-Agent string for browser requests

Best Practices

1. Leverage Zero-Config Setup - BrowserMCP runs directly via npx with no installation step. Use this for quick automation tasks where setting up a full Playwright or Puppeteer project would be overkill.

2. Chain Actions Sequentially - Browser actions depend on page state. Always wait for navigation to complete before attempting clicks or text input. Use the screenshot tool to verify the page loaded correctly between steps.

3. Use CSS Selectors for Precision - When targeting specific elements, prefer CSS selectors over text descriptions for deterministic results. Fall back to text-based targeting only when elements lack unique selectors.

4. Keep Sessions Short - Long-running browser sessions consume memory. For batch processing across many pages, break work into smaller sessions and restart the browser periodically.

5. Test with Headed Mode First - During development, disable headless mode to visually observe what the automation is doing. This catches issues like pop-ups, cookie banners, or unexpected redirects that are invisible in headless mode.

Common Issues

1. npx Download Timeout - On first run, npx downloads the package. If your network is slow, this may time out. Pre-install with npm install -g @browsermcp/mcp to avoid runtime download delays.

2. Page Not Fully Loaded - Single-page applications may report ready before all content renders. Use the screenshot tool to confirm visual state before interacting with dynamically loaded elements.

3. Permission Denied on Chromium - Some systems restrict Chromium sandbox access. If the browser fails to start, try running with the --no-sandbox flag or ensure your user has appropriate permissions.