MCP Server Directory — 104 Tested MCP Servers in 10 Categories | Skills and Servers
Directory · 10 categories · 104 servers

MCP Server Directory

The definitive directory of Model Context Protocol servers — the standard interface that lets an AI assistant safely read from and write to your actual tools and data. Every server here works with Claude Desktop, Claude Code, Cursor, Cline, Continue, and any other MCP-compliant client. Every entry has been installed, tested against real accounts, and documented by our editorial team.

104
Servers tested
10
Categories
6
Providers supported
100%
Human-tested

1What are MCP Servers?

MCP (Model Context Protocol) is the standard interface that lets AI assistants safely access external tools and data. Instead of every AI product building custom integrations for GitHub, Slack, Postgres, and everything else, MCP defines one interface that both sides speak. Servers implement it; clients (AI assistants) consume it. The result is provider-agnostic composition — the same server works with Claude, GPT-5, Gemini, and any other MCP-compliant client.

The value is controlled, scoped access to real systems. An MCP server is the middleware between an AI assistant and something you care about (your database, your inbox, your file system). It handles auth, respects rate limits, enforces safe defaults (read-only where appropriate), and gives you an audit trail. This is what turns an AI assistant from a chat tool into a genuine participant in your work.

Every server in this directory has been installed, tested against real accounts, and documented by our editorial team. Each server page includes: what it does, install commands for every major client, auth setup notes, provider-specific quality notes, and troubleshooting for common errors. All 104 servers are open-source and free to install.

2All 10 categories

Every servers in the directory lives in one of these categories. Click any card to browse.

3Editor-featured picks

Six servers that our team reaches for most often across all categories.

5How to use MCP servers

  1. Pick a category that matches something you already use daily — the value compounds when the AI can access tools you're in every day.
  2. Read the server's dedicated page for install instructions specific to your MCP client (Claude Desktop, Cursor, etc.).
  3. Set up auth with minimum necessary scope. Read-only where possible; specific-repo or specific-schema scoping over wide-open access.
  4. Test with the verification checklist on the server's page before pointing AI workflows at real data.
  5. Combine multiple servers for cross-tool workflows — the AI can compose across every registered server in one conversation. Our Custom Workflows Skills are the best starting point for multi-server orchestration.

6Frequently asked questions

An MCP (Model Context Protocol) server is a small piece of software that runs on your machine or a server and gives an AI assistant safe, scoped access to a specific tool or data source — a database, GitHub, your Slack, your file system. Without MCP servers, an AI can only respond based on what you paste into the chat. With MCP servers, the AI can query real systems, read live data, and (with permission) take actions on your behalf. This is the difference between an AI that generates suggestions and an AI that participates in your actual workflow.
Yes. Every server in this directory follows the Model Context Protocol, which means it works with Claude Desktop, Claude Code, Cursor, Cline, Continue, Warp, and any other MCP-compliant client. Setup steps differ per client (config file location, syntax) but the server itself is identical. Individual server pages have per-client install instructions.
Yes with the right setup. Every server in our directory defaults to read-only mode where possible. Write operations require explicit opt-in. For production databases, follow our Safe Read-Only MCP Setup guide. For GitHub, follow our GitHub MCP Auth Setup guide. The core rule: dedicated identity per MCP, minimum-necessary scope, audit logging on.
The MCP servers themselves are free and open-source. You may need to pay for the underlying service the server connects to (Perplexity API, Google Workspace, Salesforce, etc.) at whatever the service's own pricing is. Servers that wrap free services (filesystem, SQLite, Weather via Open-Meteo) require no payment beyond your existing tools.
Start with what you already use daily. If you're a developer: GitHub MCP and Filesystem MCP. If you're in operations: Slack MCP and Gmail MCP. If you work with data: Postgres MCP or Google Sheets MCP. Every category page shows featured picks so you can see the highlights fast.

Hit an error setting up a server?

Our sister sites cover errors, workarounds, and diagnostic recipes across the AI ecosystem.

Visit AI Error Hub

Share with