MCP Server Directory: 50+ Ready-to-Use Integrations (2026)

When I first discovered MCP servers about two months ago, I spent an embarrassing amount of time hunting through GitHub repos, random blog posts, and X threads just to figure out what was available. The Model Context Protocol ecosystem had exploded, but finding a comprehensive list? That was like searching for a specific grain of sand on a very large beach.

So I built this directory. Not because the world needed another list—it’s because the world needed an organized list.

Here’s what you’re getting: 50+ MCP servers, organized by what they actually do, with installation commands you can copy-paste, and honest recommendations about which ones are worth your time. Whether you’re a developer looking to supercharge your coding workflow or someone who just wants Claude to read their Google Drive, you’ll find what you need here.

Let’s dive into the MCP server directory that I wish existed when I started.

What Is MCP and Why This Directory Matters

If you’re new to this, MCP (Model Context Protocol) is an open standard created by Anthropic that lets AI systems like Claude connect to external tools and data sources. Think of it as a universal adapter that allows Claude to actually do things instead of just talking about them.

MCP servers are the magic behind this. Each server is a small program that exposes specific capabilities—reading files, querying databases, browsing the web, managing your tasks. When you install an MCP server for Claude Desktop, you’re essentially giving Claude new superpowers.

Here’s the thing: as of early 2026, there are over 500 community-built MCP servers floating around. That’s incredible for the ecosystem, but terrible for you trying to find the good ones. Quality varies wildly. Some servers are polished, well-documented, and actively maintained. Others were someone’s weekend project that hasn’t been updated since November 2024.

This directory does the curation work for you. I’ve tested dozens of these, read through the GitHub issues, and organized everything by use case. You’ll find:

Official servers maintained by Anthropic (always start here)
Community favorites with strong adoption and active maintenance
Specialized tools for specific workflows
Quick-start recommendations based on what you’re trying to do

One important note: this directory focuses on servers compatible with Claude Desktop as of January 2026. Claude.ai web users—you’ll need Claude Desktop for most of these. It’s free, so grab it if you haven’t.

Official Anthropic MCP Servers

Let’s start with the foundation. Anthropic maintains several official MCP servers in the modelcontextprotocol/servers repository. These are your safest bet—they’re well-documented, regularly updated, and work reliably with Claude Desktop.

The Core Six

Server	What It Does	Install Command
filesystem	Read, write, and navigate local files	`npx @modelcontextprotocol/server-filesystem /path/to/allow`
fetch	Make HTTP requests to any URL	`npx @modelcontextprotocol/server-fetch`
puppeteer	Browser automation and web scraping	`npx @modelcontextprotocol/server-puppeteer`
postgres	Query PostgreSQL databases	`npx @modelcontextprotocol/server-postgres`
github	Manage repos, issues, PRs, and files	`npx @modelcontextprotocol/server-github`
slack	Send messages and read channels	`npx @modelcontextprotocol/server-slack`

My recommendation? Start with filesystem and fetch. They’re the foundation of most useful Claude workflows. The filesystem server lets Claude read your project files, and fetch lets it pull data from any API or website.

When to Use Official vs Community Servers

Here’s my honest take: always prefer official servers when they exist. They’re maintained by the people who know MCP best, and they won’t suddenly break after a random Tuesday update.

That said, Anthropic can’t build everything. For specialized tools like Notion, Todoist, or industry-specific integrations, you’ll need community servers. The key is checking:

Last update date - Anything older than 3 months? Be cautious.
GitHub stars - Not everything, but servers with 100+ stars usually work.
Open issues - Check if people are reporting problems.
Documentation - If the README is sparse, expect headaches.

Now let’s explore what the community has built.

Productivity & Workflow Servers

These servers connect Claude to the apps you use every day. If you’re spending hours moving between Notion, Slack, and your calendar, these integrations can save serious time.

Featured Productivity Servers

1. Notion MCP

Purpose: Read and write pages, query databases, create content
GitHub: Various implementations (check mcp.so for latest)
Best for: Knowledge workers managing docs and databases
Setup complexity: Medium (requires API key from Notion)

2. Google Drive MCP

Purpose: Search, read, and organize Drive files
Best for: Teams using Google Workspace
Setup complexity: Medium (OAuth required)

3. Todoist MCP

Purpose: Create, update, and manage tasks
Best for: Task management power users
Setup complexity: Easy (API token only)

4. Linear MCP

Purpose: Manage issues, projects, and cycles
Best for: Engineering teams using Linear
Setup complexity: Easy

5. Calendar MCP (Google/Outlook)

Purpose: View, create, and manage events
Best for: Scheduling and meeting management
Setup complexity: Medium (OAuth)

6. Discord MCP

Purpose: Send messages, read channels, manage servers
Best for: Community managers and Discord power users
Setup complexity: Medium (bot token required)

My Top Pick for Productivity

Honestly? The Notion MCP has changed how I work. I use Notion for everything—project notes, research, client briefs—and being able to tell Claude “find that document about X and summarize it” is genuinely useful.

Here’s a sample configuration for your claude_desktop_config.json:

{
  "mcpServers": {
    "notion": {
      "command": "npx",
      "args": ["mcp-server-notion"],
      "env": {
        "NOTION_API_KEY": "your-notion-integration-token"
      }
    }
  }
}

The catch? You need to set up a Notion integration first (Notion Settings → Integrations → Create Integration) and share specific pages with it. Not as seamless as I’d like, but the payoff is worth the five-minute setup.

Development & Code Servers

This is where MCP really shines. Developers have built some incredibly useful integrations for code-related workflows.

Essential Developer Servers

1. GitHub MCP (Official)

Purpose: Complete GitHub integration—repos, PRs, issues, file management
Install: npx @modelcontextprotocol/server-github
Why it’s essential: Claude can read your codebase, review PRs, create issues
Setup: Requires GITHUB_PERSONAL_TOKEN environment variable

2. GitLab MCP

Purpose: Same as GitHub but for GitLab users
Best for: Enterprise teams on GitLab
Note: Community maintained, less polished than official GitHub

3. Docker MCP

Purpose: Manage containers, images, networks
Best for: DevOps workflows
Caution: Be careful with production environments

4. Kubernetes MCP

Purpose: Interact with K8s clusters
Best for: Teams managing Kubernetes deployments
Setup complexity: High (kubeconfig required)

5. npm/pip MCP

Purpose: Search packages, check dependencies
Best for: Developers researching libraries
Setup complexity: Easy

6. Sentry MCP

Purpose: Access error tracking and debugging info
Best for: Teams using Sentry for monitoring
Pro tip: Great for debugging production issues

7. VS Code MCP

Purpose: Integrate with VS Code workspace
Best for: Developers who live in VS Code
Note: Still experimental in early 2026

8. Vercel MCP

Purpose: Manage deployments, check logs
Best for: Frontend developers on Vercel
Setup complexity: Easy (API token)

The Developer Starter Pack

If you’re a developer just getting started with MCP, here’s my recommended combo:

filesystem (official) - Read your codebase
github (official) - Manage repos and PRs
fetch (official) - Hit APIs for testing
Docker MCP - Container management

This covers 80% of what I do daily. Add database servers as needed.

Data & Database Servers

Claude plus database access equals powerful data analysis without writing SQL (well, Claude writes the SQL for you).

Database Server Options

1. PostgreSQL MCP (Official)

Install: npx @modelcontextprotocol/server-postgres
Best for: Most production databases
Setup: Connection string in env

2. MySQL MCP

Purpose: Query MySQL/MariaDB databases
Community maintained
**Similar setup to Postgres

3. SQLite MCP

Purpose: Local database queries
Best for: Development, local apps, testing
Pro: No server required

4. MongoDB MCP

Purpose: NoSQL document database queries
Best for: Teams using MongoDB
Note: Query syntax differs from SQL servers

5. Redis MCP

Purpose: Key-value store operations
Best for: Caching, session management
Use case: Debugging production Redis

6. Supabase MCP

Purpose: Full Supabase access (DB + Auth + Storage)
Best for: Supabase-based applications
Nice feature: Includes auth and storage, not just DB

7. Pinecone MCP

Purpose: Vector database operations
Best for: RAG applications, semantic search
Links to: Vector databases explained

8. GraphQL MCP

Purpose: Query any GraphQL API
Best for: Modern API consumption
Flexible: Works with any GraphQL endpoint

Choosing the Right Database Server

Here’s my decision tree:

Production Postgres? → Official PostgreSQL MCP
Local development? → SQLite MCP (simple, fast)
Building RAG apps? → Pinecone or similar vector DB
Using Supabase? → Supabase MCP (all-in-one)

⚠️ Security warning: Be extremely careful connecting MCP to production databases. Always use read-only credentials when possible. Claude can execute queries, which means it can also accidentally delete things.

File & Content Management Servers

Beyond the official filesystem server, there are specialized options for different file types and storage systems.

File Management Options

1. Filesystem MCP (Official)

Install: npx @modelcontextprotocol/server-filesystem /path/to/allow
Key feature: You specify exactly which directories Claude can access
Security: Won’t access files outside allowed paths

2. S3 MCP

Purpose: Read/write to AWS S3 buckets
Best for: Cloud file storage workflows
Requires: AWS credentials configured

3. PDF MCP

Purpose: Extract text, analyze PDFs
Best for: Document processing workflows
Use case: Summarize contracts, extract data from reports

4. Image MCP

Purpose: Basic image processing
Best for: Bulk image operations
Note: Claude 4 has native vision, but this helps with processing

5. Markdown MCP

Purpose: Parse and manipulate markdown files
Best for: Documentation workflows
Links with: Filesystem for full doc management

6. Obsidian MCP

Purpose: Access Obsidian vault contents
Best for: Obsidian users wanting AI in their PKM
Setup: Point to vault directory

Security Considerations for File Access

I want to be direct about this: file access MCP servers are powerful, which means they’re also potentially dangerous. Here’s how to use them safely:

Limit directory access - Only allow specific folders, never your entire home directory
Avoid sensitive directories - Don’t point at .ssh, credential files, or secrets
Use read-only when possible - Some servers support read-only modes
Audit what Claude does - Check logs if you’re paranoid (healthy paranoia here)

Search & Research Servers

Give Claude the ability to search the web and access current information. Essential for research tasks and staying up-to-date.

Search Server Options

1. Brave Search MCP

Purpose: Web search via Brave’s privacy-focused engine
Setup: Requires Brave Search API key (free tier available)
My take: Best balance of quality and privacy

2. Exa MCP

Purpose: AI-native semantic search
Best for: Research-heavy workflows
Different from Google: Understands meaning, not just keywords

3. Perplexity MCP

Purpose: AI-powered research with citations
Best for: Fact-checking, deep research
Note: Requires Perplexity API access

4. YouTube MCP

Purpose: Search videos, fetch transcripts
Best for: Video research, content summarization
Killer feature: Full transcript access

5. Wikipedia MCP

Purpose: Direct Wikipedia access
Best for: Factual queries, definitions
Simple but useful

6. Serper MCP

Purpose: Google Search API wrapper
Best for: When you specifically need Google results
Requires: Serper API key

Combining Search for Comprehensive Research

Here’s a power move: combine multiple search servers. Have Claude check both Brave and Exa for comprehensive coverage, then use YouTube for video sources. Different search engines have different strengths.

My research setup:

brave-search for general web queries
youtube for tutorials and video content
exa for semantic/conceptual searches

AI & Machine Learning Servers

These servers let Claude work with other AI systems. Sounds meta, but it’s useful for specialized tasks that benefit from different models.

AI Integration Servers

1. Hugging Face MCP

Purpose: Access thousands of open models
Best for: Specialized ML tasks
Use case: Run sentiment analysis, embeddings, classification

2. Replicate MCP

Purpose: Run any model from Replicate’s catalog
Best for: Image generation, specialized models
Payment: Pay-per-use to Replicate

3. OpenAI MCP

Purpose: Access GPT models from Claude
Sounds weird? Useful for comparing outputs or using specialized GPTs
Rare use case but available

4. Embeddings MCP

Purpose: Generate and manage vector embeddings
Best for: RAG applications
Often paired with: Vector database servers

5. Stability MCP

Purpose: Image generation via Stability AI
Best for: Creating images within Claude conversations
Quality: Competitive with DALL-E

When to Use AI Servers with Claude

Honestly? You probably don’t need these unless you’re building something specific. Claude 4 is extremely capable on its own.

Use cases where these make sense:

Embeddings - Building RAG or semantic search systems
Image generation - Adding visual content to workflows
Specialized models - Tasks that benefit from fine-tuned models

Creative & Media Servers

For creative professionals who want AI assistance in their workflows.

Creative Server Options

1. DALL-E MCP

Purpose: Generate images from descriptions
Best for: Quick concept art, illustrations
Requires: OpenAI API key

2. Midjourney MCP

Purpose: Interface with Midjourney (via Discord API)
Best for: Higher quality art generation
Setup: Complex (Discord bot required)

3. Figma MCP

Purpose: Read and modify Figma designs
Best for: Designers working in Figma
Still experimental

4. Canva MCP

Purpose: Create and edit Canva designs
Best for: Quick graphic creation
API access: Requires Canva for Teams

5. YouTube MCP

Purpose: Already covered, but essential for video creators
For creators: Research topics, analyze competition

Creative Workflow Example

Here’s how I use creative servers:

Research topic with brave-search and youtube
Generate concept images with stability or dall-e
Draft content with Claude
Pull reference designs from figma (if applicable)

It’s not a replacement for creative tools, but it speeds up the ideation phase significantly.

Business & Enterprise Servers

Professional tools for business workflows. Enterprise users, pay attention here.

Business Integration Servers

1. Salesforce MCP

Purpose: Access CRM data, manage contacts and deals
Best for: Sales and account management teams
Setup: Requires Salesforce API access

2. HubSpot MCP

Purpose: Marketing, sales, and CRM operations
Best for: Teams using HubSpot ecosystem
Good for: Lead management, email campaigns

3. Stripe MCP

Purpose: View transactions, manage subscriptions
Best for: Finance and ops teams
Caution: Be extremely careful with payment data

4. Zapier MCP

Purpose: Trigger Zapier workflows from Claude
Best for: Automation fans
Power move: Chain multiple automations

5. Jira MCP

Purpose: Manage issues, sprints, projects
Best for: Engineering and product teams
Similar to: Linear MCP

6. Google Analytics MCP

Purpose: Access web analytics data
Best for: Marketing teams, reporting
Insight: Ask Claude to analyze trends

Enterprise Security Considerations

If you’re in an enterprise environment, talk to your security team before deploying MCP servers. Some considerations:

Data access - MCP servers can access sensitive business data
Audit logging - Ensure you can track what Claude accessed
Credential management - Use secrets managers, not env files
Network isolation - Consider where servers run

For more on enterprise deployment, check the Anthropic enterprise documentation.

How to Install MCP Servers

Let’s get practical. Here’s how to actually install and configure MCP servers for Claude Desktop.

Prerequisites

Before you start:

Claude Desktop - Download from claude.ai (free)
Node.js - Version 18+ recommended
A terminal - Mac Terminal, Windows PowerShell, or Linux shell

Step 1: Find Your Config File

Claude Desktop reads MCP config from:

Mac: ~/Library/Application Support/Claude/claude_desktop_config.json
Windows: %APPDATA%\Claude\claude_desktop_config.json
Linux: ~/.config/Claude/claude_desktop_config.json

If the file doesn’t exist, create it.

Step 2: Add Server Configuration

Here’s a template showing multiple servers:

{
  "mcpServers": {
    "filesystem": {
      "command": "npx",
      "args": [
        "@modelcontextprotocol/server-filesystem",
        "/Users/yourname/Documents",
        "/Users/yourname/Projects"
      ]
    },
    "fetch": {
      "command": "npx",
      "args": ["@modelcontextprotocol/server-fetch"]
    },
    "github": {
      "command": "npx",
      "args": ["@modelcontextprotocol/server-github"],
      "env": {
        "GITHUB_PERSONAL_TOKEN": "ghp_your_token_here"
      }
    }
  }
}

Step 3: Restart Claude Desktop

After editing the config:

Quit Claude Desktop completely
Reopen Claude Desktop
Look for the hammer icon (🔨) in the bottom-left
Click it to see available tools from your servers

Step 4: Test Your Installation

Try a simple command:

For filesystem: “List the files in my Documents folder”
For fetch: “Fetch the content from httpbin.org/json”
For github: “List my GitHub repositories”

Troubleshooting Common Issues

Server not appearing?

Check JSON syntax (trailing commas break it)
Verify the command exists (npx should work)
Check Claude Desktop logs

Permission errors?

Ensure the server has access to required paths
Check API token permissions

“Module not found” errors?

Run npm cache clean --force
Delete node_modules and reinstall

Choosing the Right Servers for Your Workflow

With 50+ options, where do you start? Here’s my guidance based on common use cases.

By Role

Writers & Content Creators:

filesystem (docs access)
brave-search (research)
notion (notes/drafts)
youtube (video research)

Software Developers:

github (code management)
filesystem (project files)
postgres/sqlite (database)
docker (containers)

Researchers:

brave-search + exa (web search)
youtube (video content)
pdf (document analysis)
notion (notes)

Business Users:

slack (communication)
calendar (scheduling)
salesforce/hubspot (CRM)
google-drive (files)

Recommended Starter Packs

Beginner Pack (3 servers):

filesystem - Local file access
fetch - Web requests
brave-search - Web search

Developer Pack (5 servers):

filesystem - Code access
github - Repo management
postgres - Database
docker - Containers
fetch - APIs

Productivity Pack (4 servers):

notion - Notes/docs
slack - Team communication
calendar - Scheduling
todoist - Tasks

Start small. Each server adds complexity and uses context window. I’d say 5-7 servers is a sweet spot for most people.

Frequently Asked Questions

How many MCP servers can I run at once?

There’s no hard limit, but practical limits exist. Each active server uses some context window capacity for its tool definitions. I run about 8 regularly without issues. If you notice Claude struggling, try reducing active servers.

Are MCP servers safe to use?

The official Anthropic servers are well-tested and safe. Community servers vary. Always review the source code for community servers before trusting them with sensitive data. Never give MCP servers access to credentials, SSH keys, or production databases without careful consideration.

How do I update an MCP server?

For npm-based servers, they update automatically when you restart Claude Desktop (npx fetches the latest). For manual installations, pull the latest from GitHub and rebuild. Check server repos for update instructions.

Can I build my own MCP server?

Absolutely! That’s the beauty of the open protocol. Check out our guide on building your first MCP server for a complete tutorial. SDKs are available for Python, TypeScript, Java, and Go.

Do MCP servers work with Claude.ai web?

As of January 2026, MCP servers only work with Claude Desktop, not the Claude.ai web interface. Anthropic has hinted at web support coming, but no timeline yet.

What’s the difference between official and community servers?

Official servers (in the modelcontextprotocol/servers repo) are maintained by Anthropic. They’re well-documented, regularly updated, and work reliably. Community servers are maintained by individuals—quality varies. Check last update date, stars, and issues before trusting community servers.

Where can I find more MCP servers?

Best directories as of 2026:

mcp.so - Community directory
GitHub search - Raw search
MCP Market - Growing marketplace

Your MCP Server Journey Starts Here

We’ve covered 50+ servers across 8 categories—from official Anthropic fundamentals to specialized business tools. The MCP ecosystem is growing fast, and this directory will help you navigate it without the hours of research I went through.

My advice? Start with the official servers. Master filesystem, fetch, and one or two others that match your workflow. Once you’re comfortable, explore the community servers that solve your specific problems.

MCP is still early. By the end of 2026, I expect we’ll have 1,000+ quality servers covering nearly every imaginable integration. The companies building tools today are racing to add MCP support—it’s becoming the standard for AI-to-tool connectivity.

Want to go deeper? Learn how to build your own MCP server and contribute to this growing ecosystem. The best server for your unique workflow might be one you create yourself.

Happy integrating.