Where can I find MCP servers to install?

The three main registries are Smithery (largest, with reviews and install instructions), mcpt (curated, quality-focused), and OpenTools (growing catalog). You can also find MCP servers on GitHub by searching for 'mcp-server' in the repository name.

Are MCP servers free?

Most MCP servers are free and open source. Some premium servers or hosted versions charge subscription fees, typically $10-50 per month. The underlying protocol is open and free. You can always build your own server for any use case at no cost beyond development time.

How do I install an MCP server?

Most MCP servers install through a JSON configuration file in your AI application. For Claude Desktop, you edit the claude_desktop_config.json file and add the server details. Many servers on Smithery include one-click install instructions. No coding is required for most installations.

Can I use multiple MCP servers at once?

Yes, and that is where the real power is. You can connect as many MCP servers as you want simultaneously. Claude or other AI assistants will automatically choose which server and tool to use based on your request. Combining calendar, email, and task servers lets you make complex requests like 'prep me for tomorrow's meetings' in a single message.

How do I know if an MCP server is safe to use?

Check the source code (most are open source on GitHub), read reviews on Smithery, and look at the permissions it requests. MCP servers run locally by default, so your data stays on your machine. Be cautious with servers that require API keys to external services — understand what data flows where before connecting.

What Are the Best MCP Servers Available Right Now?

Q: What are the best MCP servers?

The best MCP servers depend on your use case. For productivity, Google Calendar and Gmail MCP servers are essential. For development, the GitHub and filesystem servers are most popular. For research, web search and Firecrawl stand out. For business, Slack and Notion integrations are widely used. Smithery is the largest registry for browsing and installing servers.

There are thousands of MCP servers on public registries. Most of them are mediocre. Here are the ones actually worth installing, organized by what you’re trying to do.

I use MCP servers daily — this list is based on real usage, not registry star counts.

Productivity

These are the “connect these first” servers. They give AI access to the tools you already live in.

Server	What It Does	Why It’s Good
Google Calendar	Read/write calendar events, find free time, check conflicts	The single most useful first MCP server. “What’s on my calendar tomorrow?” becomes instant.
Gmail	Search, read, draft, and organize email	”Find all emails from [person] this week” saves real time. Draft responses without switching apps.
Slack	Read channels, send messages, search history	Let AI summarize what you missed overnight. Post updates without opening Slack.
Notion	Read and write to Notion databases and pages	If Notion is your second brain, this gives AI direct access to it.
Todoist / Linear	Task management — create, update, query tasks	”What’s overdue?” and “Add a task to follow up with Sarah” from any AI chat.
Google Drive	Search and read documents, sheets, slides	Access your files without manual exports.

Stack these: Calendar + Gmail + Todoist gives AI enough context to say “You have a meeting with Sarah at 2pm, here’s her last email, and you have an overdue task to send her the proposal.” One question, three servers working together.

Development

If you write code, these are essential.

Server	What It Does	Why It’s Good
GitHub	Repos, PRs, issues, code search, actions	The most-installed dev MCP server. Manage repos from your AI chat.
Filesystem	Read, write, search local files	Let AI access your project files directly. Foundation for code-aware assistance.
PostgreSQL / SQLite	Query databases, inspect schemas	”Show me users who signed up this week” without writing SQL manually.
Docker	Manage containers, images, logs	Check container status and read logs from your AI chat.
Sentry	Error tracking, issue search	”What errors are trending in production?” with real data.

The power combo: GitHub + Filesystem + Database lets AI understand your codebase, check live data, and create PRs — all from natural language requests.

Research & Information

These extend AI’s knowledge beyond its training data.

Server	What It Does	Why It’s Good
Web Search	Real-time web search with page content extraction	AI can look things up instead of guessing. Essential for current events and recent information.
Firecrawl	Full website scraping, crawling, structured extraction	Extract data from any website. Scrape documentation, product pages, competitor sites.
PubMed	Search academic medical literature	Search articles, get full text, find related research. Invaluable for medical and science professionals.
Wikipedia	Structured Wikipedia access	Quick factual lookups with source citations.
Brave Search	Privacy-focused web search	Alternative to Google-based search servers. Good results, no tracking.

My daily use: Web Search + Firecrawl handles 90% of research tasks. AI searches, reads the pages, and synthesizes a brief. What used to take 30 minutes of tab-hopping takes one prompt.

Business & Operations

For running your business through AI.

Server	What It Does	Why It’s Good
Stripe	Payment data, subscriptions, invoices	”How much revenue did we do last month?” with real numbers.
HubSpot / Salesforce	CRM access — contacts, deals, activities	Prep for sales calls without digging through the CRM manually.
QuickBooks	Financial data, invoicing	Ask AI about your books in plain language.
Shopify	Store data, orders, products, analytics	Monitor your store from any AI chat.
Airtable	Flexible database access	If you run operations through Airtable, this unlocks AI access to all of it.

Creative & Content

For creators, writers, and marketers.

Server	What It Does	Why It’s Good
Figma	Read designs, get screenshots, generate diagrams	Bridge the gap between design and code. AI can see your Figma files.
Canva	Create and edit designs, manage assets	Generate designs from descriptions, edit existing work.
Substack	Manage newsletter drafts, stats, subscribers	Monitor your newsletter performance from AI chat.
YouTube	Search videos, get transcripts, channel data	Research video content, pull transcripts for analysis.
Unsplash / Pexels	Search and download stock photos	Find images for content without leaving your workflow.

Where to Find and Install Them

The Three Main Registries

Registry	Strength	Best For
Smithery	Largest catalog, reviews, one-click install instructions	Browsing and discovering new servers
mcpt	Curated, quality-focused	Finding well-maintained, reliable servers
OpenTools	Growing catalog, good categorization	Alternative browsing when Smithery doesn’t have what you need

How to Install (Claude Desktop)

Most MCP servers follow the same pattern:

Find the server on a registry or GitHub
Open your Claude Desktop config file (claude_desktop_config.json)
Add the server entry (usually 3-5 lines of JSON)
Restart Claude Desktop

That’s it. No coding, no build steps for most servers. The registry listings include copy-paste config snippets.

How to Install (Claude Code / CLI)

If you use Claude Code in the terminal:

Add servers to your project’s .mcp.json or global settings
Claude Code picks them up automatically on next start
Use claude mcp commands to manage servers

How I Actually Use These

My daily Claude Code session connects to:

Slack — read channels, post updates
Gmail — search and draft emails
Google Calendar — check schedule, find conflicts
Figma — pull design context into code
Web Search — real-time lookups
Firecrawl — deep web scraping and research
GitHub — manage repos and PRs
Filesystem — full access to project files

On any given day, I might say: “Check my calendar for tomorrow, find the Figma file for the project I’m meeting about, and set up a new branch in the GitHub repo.” Three servers, one sentence, done in seconds.

My OpenClaw gateway (the local AI automation layer) also uses MCP servers through its cron jobs — web search for research pipelines, filesystem for log analysis, database access for metrics. These run 26 automated tasks daily, all locally.

How to Pick Your First Server

Don’t overthink it. Pick based on what you use most:

If you spend time in…	Install this first
Google Calendar	Google Calendar MCP
Email	Gmail MCP
Slack	Slack MCP
GitHub	GitHub MCP
Web research	Web Search MCP
Notion	Notion MCP

Start with one. Use it for a week. Then add a second one. The compounding effect is real — each new server makes all the others more useful because AI can combine them.

Frequently Asked Questions

How many MCP servers can I run at once?

As many as you want. Claude handles multiple servers simultaneously and chooses which to call based on your request. I regularly run 8+ at the same time.

Do MCP servers slow down my AI?

No. Servers only activate when the AI decides to call them. Having 10 servers connected doesn’t affect response time for questions that don’t need them.

Can I build my own MCP server?

Yes. The protocol is open, and building a basic server takes a few hours if you can write JavaScript/TypeScript or Python. There are starter templates on GitHub. I’ll cover this in a future article.

Are there MCP servers for [specific tool]?

Probably. Search Smithery first — if it exists, it’s likely listed there. If not, check GitHub for “[tool name] mcp server.” The ecosystem is growing fast and new servers appear daily.

This is part of the ASTGL Definitive Answers series — structured, practical answers to the questions people actually ask about AI automation, MCP servers, and local AI infrastructure.