Finding MCPWhat is mcp?Model Context Protocol - a standard that lets AI tools connect to external services like databases, issue trackers, or APIs. Ask AI for more servers

MCP servers are distributed as regular packages, mostly npm packages or standalone executables. Here are the main places to find them:

• Anthropic's official directory: A curated list of verified servers at the MCP documentation site
• npm registryWhat is registry?A server that stores and distributes packages or container images - npm registry for JavaScript packages, Docker Hub for container images. Ask AI for more: Search for @modelcontextprotocol/server-* for official servers, or browse community packages tagged with mcp
• GitHub: Many developers publish MCP servers as open-source repos
• Community lists: Curated "awesome-mcp" lists on GitHub with categorized servers

The official servers maintained by Anthropic cover the most common use cases: filesystem access, GitHub, PostgreSQL, Slack, Google Drive, and more.

Configuring servers in Claude Desktop

Claude Desktop reads its MCPWhat is mcp?Model Context Protocol - a standard that lets AI tools connect to external services like databases, issue trackers, or APIs. Ask AI for more configuration from a JSONWhat is json?A text format for exchanging data between systems. It uses key-value pairs and arrays, and every programming language can read and write it. Ask AI for more file called claude_desktop_config.json. The location depends on your operating system:

• macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
• Windows: %APPDATA%\Claude\claude_desktop_config.json

Here is a basic configuration with a filesystem server:

{
  "mcpServers": {
    "filesystem": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-filesystem", "/Users/you/projects"],
      "env": {}
    }
  }
}

Let's break this down:

• "filesystem", the name you give this server (can be anything descriptive)
• "command", the command to start the server (npx runs it directly from npm)
• "args", arguments passed to the command: -y auto-confirms the install, the package name, and the directory to expose
• "env", environment variables the server needs (empty here, but used for APIWhat is api?A set of rules that lets one program talk to another, usually over the internet, by sending requests and getting responses. Ask AI for more keys)

After saving this file and restarting Claude Desktop, you will see a hammer icon in the input bar indicating MCP tools are available.

Configuring servers in Claude Code

Claude Code has a built-in command for adding MCPWhat is mcp?Model Context Protocol - a standard that lets AI tools connect to external services like databases, issue trackers, or APIs. Ask AI for more servers:

# Add a filesystem server
claude mcp add filesystem -- npx -y @modelcontextprotocol/server-filesystem /Users/you/projects

# Add a GitHub server with an environment variable
claude mcp add github -e GITHUB_TOKEN=ghp_xxxxxxxxxxxx -- npx -y @modelcontextprotocol/server-github

# List configured servers
claude mcp list

# Remove a server
claude mcp remove filesystem

The claude mcp add syntax is: claude mcp add <name> [flags] -- <command> [args...]. The -- separates Claude Code flags from the server command.

Common MCPWhat is mcp?Model Context Protocol - a standard that lets AI tools connect to external services like databases, issue trackers, or APIs. Ask AI for more servers

Here are the most widely used MCP servers and what they provide:

A multi-server configuration

You can configure multiple servers at once. Here is a realistic developer setup:

{
  "mcpServers": {
    "project-files": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-filesystem", "/Users/you/projects/my-app"]
    },
    "github": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-github"],
      "env": {
        "GITHUB_TOKEN": "${GITHUB_TOKEN}"
      }
    },
    "database": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-postgres"],
      "env": {
        "DATABASE_URL": "postgres://readonly_user:password@localhost:5432/myapp_dev"
      }
    }
  }
}

With this configuration, Claude can read your project files, check your GitHub issues and PRs, and query your development database, all in one conversation.

Verifying your connection

After configuring a server, verify it is working:

1. Claude Desktop: Look for the hammer icon in the input bar. Click it to see available tools. If the icon is missing, check your JSONWhat is json?A text format for exchanging data between systems. It uses key-value pairs and arrays, and every programming language can read and write it. Ask AI for more file for syntax errors.
2. Claude Code: Run claude mcp list and confirm your server appears with status "connected".
3. Test with a simple request: Ask Claude something that requires the server. For a filesystem server, ask "What files are in my project directory?" If Claude can answer, the server is connected.

Common connection errors

Security: the principle of least privilege

MCPWhat is mcp?Model Context Protocol - a standard that lets AI tools connect to external services like databases, issue trackers, or APIs. Ask AI for more servers run on your machine with your permissions. This is powerful, and dangerous if misconfigured.

Rule 1: ScopeWhat is scope?The area of your code where a variable is accessible; variables declared inside a function or block are invisible outside it. Ask AI for more access narrowly

Bad: give the filesystem server access to your entire home directory

"args": ["-y", "@modelcontextprotocol/server-filesystem", "/Users/you"]

Good: scope it to just the project you are working on

"args": ["-y", "@modelcontextprotocol/server-filesystem", "/Users/you/projects/my-app"]

Rule 2: Use read-only access when possible

If Claude only needs to read your database, connect with a read-only database user. Do not give it the admin password.

Rule 3: Avoid hardcoding secrets

Instead of putting APIWhat is api?A set of rules that lets one program talk to another, usually over the internet, by sending requests and getting responses. Ask AI for more keys directly in your config file, reference environment variables:

"env": {
  "GITHUB_TOKEN": "${GITHUB_TOKEN}"
}

Then set the variable in your shell profile (~/.zshrc or ~/.bashrc):

export GITHUB_TOKEN="ghp_your_actual_token"

Rule 4: Review third-party servers

Before installing a community MCP server, check:

• Is the source code public? Can you read what it does?
• Does it request more permissions than it needs?
• Is it actively maintained?
• Do other developers use and trust it?

Treat MCP servers like browser extensions: useful, but worth inspecting before you install.