MCP Browser Server

What is MCP Browser Server?

A Model Context Protocol (MCP) server that provides browser automation capabilities using Playwright. It enables AI assistants to interact with web pages through a standardized interface, supporting multi-browser automation (Chromium, Firefox, WebKit) with tools for navigation, clicking, typing, screenshotting, and JavaScript execution. It is designed for AI-powered web testing, scraping, and debugging workflows.

How to use MCP Browser Server?

Install dependencies via npm install, build with npm run build, and ensure Playwright browsers are installed using npx playwright install. For VS Code integration, configure the server in settings.json as an MCP server with the command node pointing to the built index.js. Once configured, any MCP-compatible AI assistant (such as Chat.fans agents or GitHub Copilot Chat) can invoke the provided tools.

Key features of MCP Browser Server

• Multi-browser support: Chromium, Firefox, and WebKit
• Comprehensive automation: navigate, click, type, screenshot, and more
• JavaScript execution in the browser context
• Element interaction: wait for elements, get text, interact with forms
• Scroll tool with direction, pixel, and behavior control
• Scrollability check tool
• AI-powered screenshot analysis using Gemma3 via Ollama (optional)
• Type-safe implementation with TypeScript and Zod

Use cases of MCP Browser Server

• Automate web application testing with AI assistants
• Perform web scraping and data extraction
• Debug web pages by capturing screenshots and console logs
• Run custom JavaScript on live pages for verification
• Analyze page structure and content with vision-capable AI models

FAQ from MCP Browser Server

What are the system requirements?

Node.js, Playwright browsers installed (via npx playwright install), and on Linux additional system dependencies (via sudo npx playwright install-deps). The AI screenshot analysis feature requires Ollama with a Gemma3 model (e.g., gemma3:4b).

How do I integrate MCP Browser Server with VS Code?

Add an MCP server entry to your VS Code settings.json using the command node and the path to the built build/index.js. After configuration, Chat.fans agents and GitHub Copilot Chat can use the browser tools.

Which browsers are supported?

Chromium, Firefox, and WebKit are all supported.

Do I need Ollama to use the server?

No. All standard browser automation tools work without Ollama. The analyze_screenshot tool is optional and only available if Ollama is installed with a vision-capable model.

Where are screenshots and console logs saved?

Screenshots can be saved to a custom path specified in the tool call. By default, generated screenshots from tests are stored in a screenshots/ directory. Console logs are returned directly to the caller.