Deploy a Production-Ready MCP Server in Under Two Minutes

The promise of Model Context Protocol (MCP) servers is enticing, but standing up a production-ready instance often feels like a multi-hour project. What if you could deploy a fully functional MCP server in under two minutes with a single click? The Naftiko Framework’s Shipyard project delivers exactly that, using Cloudflare’s containerized infrastructure to bridge the gap between local prototyping and a live, scalable endpoint.

What You’re Actually Getting

When you deploy the Shipyard MCP server, you’re not just spinning up a mock server—you’re activating a complete system with three protocol adapters, two external APIs, and a production-grade deployment architecture. The setup includes:

• MCP Adapter: Handles structured interactions via POST /mcp with endpoints like list-ships, get-ship, and create-voyage.
• REST Adapter: Exposes the same data as a standard REST API under /api/..., hosted on the same domain.
• Skill Adapter: Provides /skill endpoints for structured agent discovery and integration.

The server is pre-wired to two public mock APIs, eliminating the need for upstream credentials:

• Maritime Registry API: A bearer-authenticated mock API at mocks.naftiko.net/rest/naftiko-shipyard-maritime-registry-api/1.0.0-alpha2.
• Legacy Dockyard API: An API-key-authenticated mock API at mocks.naftiko.net/rest/naftiko-shipyard-legacy-dockyard-api/1.0.0-alpha2.

A dummy bearer token (sk-mcp-YYYYYYYYYYYY) is included in the deployment, so you can test immediately without configuring upstream access.

The Cloudflare Deployment Blueprint

The Naftiko Framework’s engine runs as a Docker container (ghcr.io/naftiko/framework:latest), orchestrated by Cloudflare Containers behind a Worker. The deployment leverages a familiar pattern:

• Durable Object Broker: Manages container lifecycle and traffic routing.
• Worker Proxy: Routes requests to the correct port based on the endpoint:
• /mcp → MCP server (port 3001)
• /skill → Skill server (port 3003)
• All other paths → REST server (port 3002)
• Five Core Files: A capability YAML, three shared imports, a Dockerfile, a Wrangler config, and a small Worker script.

This architecture isn’t just for demos—it mirrors what powers production deployments like manage-companies. The Shipyard example is a scaled-down version of the same pattern, designed to demonstrate the full lifecycle from YAML configuration to a live endpoint.

From Click to Live in Two Minutes

The deployment process is intentionally frictionless:

1. Click the deployment button in the Shipyard repository.
2. Cloudflare authenticates your account, forks the repo, and provisions a Worker with a Container.
3. The Docker image builds automatically, and a *.workers.dev hostname is assigned.

Within two to three minutes, you’ll have a fully operational MCP server with the following endpoints:

• ` – A landing page with deployment details.
• ` – The MCP server endpoint (give this to your AI tools).
• ` – A REST adapter for the same data.
• ` – A Skill server for structured agent interactions.

No local setup, no Docker configuration, and no manual API wiring—just a live endpoint ready for integration.

Integrating the MCP Server with AI Tools

Once deployed, connecting the MCP server to your AI tools is straightforward. Replace <your-worker> in the examples below with your assigned *.workers.dev subdomain. The default bearer token is sk-mcp-YYYYYYYYYYYY, unless you’ve customized it during deployment.

For Claude Desktop

Edit your configuration file to add the MCP server:

{
  "mcpServers": {
    "naftiko-shipyard": {
      "type": "streamable-http",
      "url": "
      "headers": {
        "Authorization": "Bearer sk-mcp-YYYYYYYYYYYY"
      }
    }
  }
}

Save the file and restart Claude Desktop. The MCP server tools will appear in the picker.

For Claude Code

Use the CLI to register the MCP server:

claude mcp add --transport http naftiko-shipyard  \
  --header "Authorization: Bearer sk-mcp-YYYYYYYYYYYY"

Verify the setup with:

claude mcp list

For ChatGPT (Pro/Enterprise)

Configure a custom connector via the ChatGPT interface:

1. Navigate to Settings → Connectors → Custom.
2. Add a new connector with:

• Server URL: `
• Authentication: Custom header named Authorization with value Bearer sk-mcp-YYYYYYYYYYYY.

For Gemini CLI

Update your ~/.gemini/settings.json file:

{
  "mcpServers": {
    "naftiko-shipyard": {
      "httpUrl": "
      "headers": {
        "Authorization": "Bearer sk-mcp-YYYYYYYYYYYY"
      }
    }
  }
}

Restart the Gemini CLI and confirm the connection with /mcp list.

The Real-World Impact of This Approach

The Shipyard example is more than a tutorial—it’s a blueprint for how MCP servers should be deployed in production. The same one-click pattern used here powers real-world capabilities like manage-companies in Naftiko’s infrastructure. The key insight?

You can replace the capability YAML in your fork, redeploy, and instantly spin up a new MCP server at the same URL. This isn’t just a demo; it’s a repeatable workflow for deploying production-grade MCP servers with minimal overhead.

In a landscape where AI tools increasingly rely on structured data sources, the ability to deploy and manage MCP servers in minutes—not hours or days—is a game-changer. With Cloudflare Containers and the Naftiko Framework, the barrier to entry for MCP server deployment has never been lower.