Most people use Claude like a smarter Google. You type, it talks back. That is one hand tied behind its back.
An MCP changes that. It plugs Claude into a real tool or a website, so it stops talking and starts doing. It runs your browser, scans full sites, makes images, and searches the live web. Here are five worth setting up, in the order I would add them.
An MCP is the difference between a chatbot and an operator.
01. What an MCP actually is
The plug
In easy words, an MCP is a plug that connects Claude to a tool or a website. Without one, Claude can only talk. With one, it can run your browser, scan websites, make images, and search the live web. You set it up once, and it stays available in every chat after that.
02. Before you start
Desktop or Code, plus Node
These 5 MCPs do not run on the claude.ai web version. You need one of these:
- Claude Desktop app (download at claude.ai/download), or
- Claude Code, if you are comfortable with the terminal.
You also need Node.js on your system. To check, open your terminal and type:
node -vIf a version number shows up, you are good. If you see "command not found", install it from nodejs.org.
03. Add any MCP, learn once
Two ways
There are two ways to add any MCP. Use whichever suits you.
Way 1, Claude Code. Run one command and you are done. Each MCP below has its command.
Way 2, Claude Desktop. Open the config file, paste a JSON block, restart the app. The file lives here:
- Mac:
~/Library/Application Support/Claude/claude_desktop_config.json - Windows:
%APPDATA%\Claude\claude_desktop_config.json
Every new MCP is just one more block inside mcpServers. Everything else stays the same.
After you edit the config, fully quit Claude Desktop and open it again. If you skip that, the new MCP will not load.
04. Playwright, start here
Free, no key
GitHub: github.com/microsoft/playwright-mcp
What it does: Claude controls your browser. It clicks buttons, fills forms, and automates tasks. It works on the page structure, not screenshots, so it is fast and reliable.
API key? No. It is free and the easiest one, so start here.
# Claude Code
claude mcp add playwright npx @playwright/mcp@latest// Claude Desktop config
{ "mcpServers": { "playwright": { "command": "npx",
"args": ["@playwright/mcp@latest"] } } }Try this: "Use Playwright MCP to open demo.playwright.dev/todomvc and add two todo items."
05. Firecrawl
Free key, faster research
GitHub: github.com/firecrawl/firecrawl-mcp-server
What it does: Scans a full website and brings all the content straight into your chat. Research gets much faster.
API key? Yes, but free. The free tier gives you 500 credits. Get your key at firecrawl.dev, make an account, then copy it from the API keys section.
# Claude Code
claude mcp add firecrawl npx -y firecrawl-mcp --env FIRECRAWL_API_KEY=fc-your-keyTry this: "Scrape stripe.com/pricing and extract all plan prices."
06. Glif
Images and short videos
GitHub: github.com/glifxyz/glif-mcp-server
What it does: Connects Claude to glif.app's AI workflows, which include image and video models. So you can make images and short videos right from your chat. You can run glifs, search featured workflows, and save them as tools.
API key? Yes. Get your token at glif.app/settings/api-tokens.
// Claude Desktop config
{ "mcpServers": { "glif": { "command": "npx",
"args": ["-y", "@glifxyz/glif-mcp-server@latest"],
"env": { "GLIF_API_TOKEN": "your-token-here" } } } }Try this: "Use Glif to generate a meme about AI."
07. Perplexity
Live web search
GitHub: github.com/perplexityai/modelcontextprotocol
What it does: Gives Claude live web search. Real-time results without leaving Claude. It runs on Sonar models and answers with citations.
API key? Yes, and note this one is paid. Create the key from the Perplexity API portal. If you just want to try things first, set up the free ones and add this later.
# Claude Code
claude mcp add perplexity --env PERPLEXITY_API_KEY="your_key" -- npx -y @perplexity-ai/mcp-serverTry this: "Search the web for the latest AI news from this week."
08. Chrome
Your real, logged-in browser
GitHub: github.com/ChromeDevTools/chrome-devtools-mcp
What it does: Claude sees and controls your actual Chrome browser and your open tabs. Your logins work too, so wherever you are already signed in, Claude can work directly. I recommend the official Chrome DevTools MCP: it runs on Windows and Mac and installs like the others.
API key? No. You need Chrome 144 or newer. Turn on remote debugging at chrome://inspect/#remote-debugging.
# Claude Code
claude mcp add chrome-devtools npx chrome-devtools-mcp@latestEasier, if you have Pro or Max: install Anthropic's official "Claude in Chrome" extension from the Chrome Web Store. It connects to your session and shares your login. No config needed.
Playwright vs Chrome: Playwright opens a fresh clean browser, best for automation. Chrome uses your real logged-in browser, where you are already signed in. Keep both.
09. Common problems
Quick fixes
- "command not found": Node.js is not installed. Get it from nodejs.org.
- MCP not showing up: after editing the config, fully quit Claude Desktop and reopen it.
- API key error: check the key is pasted correctly, no extra spaces.
- Something stuck: add one MCP at a time, so you know which one is running.
Start with the free ones (Playwright, Firecrawl, Chrome), then add the rest. Once these are set up, Claude stops being just a chatbot and becomes an operator.
Stuck on any setup? Comment or DM me and I will help you out.
// Free newsletter
I send out guides like this every week
Real setups, real sources, no hype. Drop your email and I'll send you the next one.