How to Host an MCP Server: A Quick-Start Walkthrough

If you already know you need to host an MCP server and just want the fastest path to a working endpoint, this is the short version. For the full breakdown of what production hosting requires and how the options compare, see our complete guide to MCP server hosting.

Step 1: Decide If You're Self-Hosting or Using a Managed Platform

This is the one decision that determines everything else. If you already run servers you trust and want full control, self-host. If you want a working endpoint in minutes without managing a reverse proxy or certificate renewal, use a managed platform.

Self-hosting checklist:

• A VPS or cloud instance
• Node.js or Python runtime installed
• Nginx (or similar) for TLS termination
• A process manager (PM2, systemd, or a container orchestrator)

Managed platform checklist:

• An account on a platform like MCPCore
• That's it. The endpoint, TLS certificate, and subdomain are provisioned automatically.

Step 2a: Self-Hosting in Five Commands

# Install Node.js
curl -o- https://raw.githubusercontent.com/nvm-sh/nvm/v0.40.0/install.sh | bash
nvm install 22

# Clone and install your server
git clone https://github.com/your-org/your-mcp-server.git
cd your-mcp-server
npm install --production

# Keep it running
npm install -g pm2
pm2 start server.js --name mcp-server
pm2 save

Then point Nginx at localhost:3000 with a valid TLS certificate from Certbot, and your endpoint is live at https://mcp.yourdomain.com/mcp. Our full deployment guide covers the Nginx config, rate limiting, and health checks in detail.

Step 2b: Managed Hosting in Three Clicks

1. Create an account and create a new server.
2. Choose an authentication mode: Public, API Key, OAuth 2.0, or Bearer Token.
3. Write your first tool in the browser IDE and save it.

Your server is reachable at https://your-subdomain.mcpcore.io/mcp as soon as it is created, before you have even written a tool.

Step 3: Test the Endpoint

Regardless of which path you took, verify the server responds before wiring it into an AI client:

curl -X POST https://your-endpoint/mcp \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'

A healthy server returns a JSON-RPC response listing its available tools. If you get a connection error, check that your TLS certificate is valid and that the server process is actually running.

Step 4: Connect an AI Client

Point Claude Desktop, Cursor, Windsurf, or another MCP client at your endpoint URL. Our guide to connecting AI clients to a custom MCP server has copy-paste configuration for each.

Step 5: Add Authentication Before You Share the URL

An endpoint with no authentication is fine for local testing and dangerous for anything shared. Before you hand the URL to anyone else, require at least an API key. See our MCP server security page for how each authentication mode works.

---

This is a condensed quick-start. For the reasoning behind each hosting option and how to choose between them, see MCP Server Hosting: The Complete Guide.