All Tutorials
AI Agents
Intermediate
15 min

Deploy an MCP Server on OSC

Deploy your custom MCP server from GitHub and get a production-ready HTTPS endpoint in minutes — all through your AI agent.

What You'll Learn

Deploy an MCP server to production with HTTPS
Configure HTTP/SSE transport for cloud hosting
Share your MCP endpoint with any AI agent
Manage deployed MCP servers on OSC

Prerequisites

  • Completed a Getting Started tutorial: Claude Code or Codex
  • A GitHub repository with an MCP server (HTTP/SSE transport)

Step-by-Step Guide

1

Prerequisites

Before you begin, make sure you have:

  • An OSC account (sign up at app.osaas.io)
  • An AI agent connected to OSC via MCP (Claude Code, Codex, or ChatGPT)
  • A GitHub repository with your MCP server code (public or private)

New to OSC? Complete the Getting Started with Claude Code tutorial first.

2

Prepare Your MCP Server

Your MCP server needs two things to work on OSC:

1. Use HTTP/SSE transport (not stdio)

OSC deploys your server as a web service with an HTTPS endpoint. Your server must accept MCP connections over HTTP with SSE, not over standard I/O.

2. Respect the PORT environment variable

OSC sets the PORT environment variable at runtime. Your server must listen on this port.

const port = process.env.PORT || 3000;
server.listen(port);

Important: MCP servers using stdio transport only (like most local-only servers) need to be updated to support HTTP/SSE before deploying to OSC.

3

Deploy via Your AI Agent

Ask your AI agent to deploy your MCP server to OSC:

> Deploy my MCP server from https://github.com/user/my-mcp-server as a Node.js app named mymcpserver

Your AI agent will use the create-my-app tool to:

  • Pull your code from GitHub
  • Build a container image
  • Deploy it to OSC infrastructure

Tip: The app name must be alphanumeric only (no hyphens or special characters). For Python servers, specify python as the runtime type.

4

Wait for Deployment

OSC will build and deploy your application. This typically takes 2-5 minutes. Your AI agent will track the progress and notify you when it's ready.

You can check the status anytime:

> What's the status of my app mymcpserver?
5

Get Your Endpoint URL

Once deployed, OSC provides a public HTTPS URL for your MCP server:

https://mymcpserver.auto.prod.osaas.io

Your MCP endpoint is typically at /mcp or the root path, depending on your server's configuration.

6

Configure MCP Clients

Share your MCP server with anyone by providing this configuration:

{
  "mcpServers": {
    "my-server": {
      "type": "http",
      "url": "https://mymcpserver.auto.prod.osaas.io/mcp"
    }
  }
}

Add this to the MCP configuration of any compatible AI agent — Claude Desktop, VS Code, Codex, or ChatGPT.

7

Verify It Works

Test your deployed MCP server by asking an AI agent to use it:

> List the tools available from this MCP server

Success! Your MCP server is now live and accessible to any MCP-compatible AI agent via HTTPS.

Your MCP Server is Live!

You've deployed your own MCP server to Open Source Cloud. Anyone with the HTTPS URL can connect their AI agent to your server — no installation, no configuration.

What's Next?

Add Configuration

Use a parameter store to manage environment variables and secrets for your MCP server.

> Set up a parameter store for my MCP server

Update Your Server

Push updates to GitHub and redeploy to get the latest version live.

> Restart my app mymcpserver

Explore More

Discover other tutorials or browse the full documentation to unlock more possibilities with Open Source Cloud.

DocumentationMore Tutorials