How to install an MCP server

To install an MCP server, add a short JSON config snippet to your AI app and restart it. Local plugins require Node.js or Python; remote plugins need only a URL. The whole process takes a few minutes.

New to MCP? Start with What is an MCP server? to understand what you are installing and why.

There are two types:

• Local plugins run on your computer. You add a short config and your AI app launches them automatically.
• Remote plugins run in the cloud. You just paste a URL: nothing to install.

Check the plugin listing on MCP Marketplace to see which type you're installing.

Installing a local plugin

What you need

• An AI app that supports MCP (Claude Desktop, Claude Code, Cursor, VS Code, Windsurf, or similar)
• Node.js 18+ (for JavaScript/TypeScript plugins) or Python 3.10+ (for Python plugins): most plugins tell you which one they need
• Any credentials the plugin requires (API key, token, etc.)

Step 1: Find the plugin you want

Browse MCP Marketplace to find what you need. Each listing shows:

• What the plugin does
• What credentials you need
• A security score
• One-click install instructions

For a curated list of the most useful servers, see Best MCP servers for developers.

Step 2: Get your credentials

If the plugin connects to a service (GitHub, Slack, Notion, etc.), you'll need credentials from that service first: usually an API key or token. The plugin listing tells you exactly what's needed and where to get it.

Keep credentials safe. Don't share them or post them publicly.

Step 3: Add it to your AI app

Each AI app has a config file where you register plugins. The listing on MCP Marketplace shows the exact config for your app.

For JavaScript/TypeScript plugins (installed via npx):

{
  "mcpServers": {
    "your-plugin-name": {
      "command": "npx",
      "args": ["-y", "@package/server-name"],
      "env": {
        "API_KEY": "your-key-here"
      }
    }
  }
}

For Python plugins (installed via uvx):

{
  "mcpServers": {
    "your-plugin-name": {
      "command": "uvx",
      "args": ["your-plugin-name"],
      "env": {
        "API_KEY": "your-key-here"
      }
    }
  }
}

Tip: GUI apps like Claude Desktop may not find uvx in your PATH. If the server won't connect, replace "uvx" with the full path — run which uvx in your terminal to find it (e.g. "/Users/you/.local/bin/uvx").

Copy the exact configuration from the plugin's install instructions and replace the placeholder values with your actual credentials.

Step 4: Restart your AI app

Save the config file and restart your AI app. It will launch the plugin automatically in the background.

Step 5: Try it out

Ask your AI assistant what tools it has available. Then try something simple: if you installed the GitHub plugin, try "List my recent GitHub repositories." If it works, you're all set.

Installing a remote plugin

Remote plugins are even simpler: there's nothing to install on your computer. The plugin runs in the cloud and your AI app connects to it over a URL.

What you need

• An AI app that supports remote MCP connections (Claude Desktop 0.10+, Cursor, VS Code, Windsurf)
• The plugin's URL (found on its listing page)
• Any required credentials

Step 1: Get the remote URL

Find the plugin on MCP Marketplace and copy the remote URL from the listing. It looks something like:

https://api.example.com/mcp

Step 2: Add it to your config

The config for a remote plugin uses url instead of command:

{
  "mcpServers": {
    "your-plugin-name": {
      "url": "https://api.example.com/mcp"
    }
  }
}

If the plugin requires an API key:

{
  "mcpServers": {
    "your-plugin-name": {
      "url": "https://api.example.com/mcp",
      "headers": {
        "Authorization": "Bearer your-api-key-here"
      }
    }
  }
}

Step 3: Restart and try it

Save the config, restart your AI app, and ask what tools are available. Remote plugins connect quickly since there's no startup delay.

Troubleshooting

Plugin doesn't appear: Check the config file for JSON errors: a missing comma or bracket will cause it to fail silently. Try pasting your config into a JSON validator.

Authentication errors: Double-check that your credentials are correct and have the right permissions.

"Command not found": For JavaScript plugins, make sure Node.js is installed (node --version in a terminal). For Python plugins, make sure the package is installed (pip install your-plugin-name).

Connection refused (remote): Confirm the URL is correct. Some remote plugins require an active subscription or API key.

Tools not showing up: Your AI app may need an update to support remote plugins. Check the app's release notes.

One-click install on MCP Marketplace

Many plugins on MCP Marketplace support one-click install, which generates the config snippet automatically for your specific AI app. Check the plugin listing for install options.

For specific server setup guides, see GitHub MCP server setup or browse the best MCP servers for developers for more options.