Build & Deploy a Remote MCP Server to Google Cloud Run in 10 Minutes

Thank you for reading my latest newsletter, "Build & Deploy a Remote MCP Server to Google Cloud Run in 10 Minutes." Here at LinkedIn, I regularly write about management and technology trends. To read my future articles, join my network here or click Follow or Subscribe to my newsletter AI Innovation.

Imagine giving your LLM-powered applications the gift of accurate context in real time while ensuring enterprise-grade scalability and security. That's what Model Context Protocol (MCP) and Google Cloud Run make possible.

Today, I’m walking you through how you can deploy a remote MCP server to Cloud Run in under 10 minutes. Why does this matter? Because modern AI agents are only as good as the context they can reason with.

Let’s break it down:

Why MCP Matters

Contextual understanding in AI has always been a challenge. Anthropic’s Model Context Protocol (MCP) bridges that gap, providing a standardized, structured way for tools, APIs, and data to feed into LLMs securely and reliably.

🧠 Think of MCP as the “middleware” between your APIs and the LLMs.

Traditionally, MCP servers were local. Now with streamable HTTP as a supported transport, you can host them remotely, enabling powerful collaboration and centralized tool management.

☁️ Why Google Cloud Run?

Google Cloud Run makes it dead simple to:

✅ Scale automatically

✅ Share across teams securely using IAM

✅ Enforce authentication (no rogue API calls!)

✅ Integrate with modern CI/CD pipelines

Running your MCP server on Cloud Run = zero ops overhead, maximum security.

⚙️ TL;DR Deployment Guide

1. Install uv for Python project management
2. Create your project & file structure
3. Build your MCP server with FastMCP (we used math tools: add and subtract)
4. Use Docker with a lightweight Python base image
5. Deploy to Cloud Run using either source or container image
6. Secure with IAM (--no-allow-unauthenticated)
7. Run Cloud Run Proxy locally for testing
8. Test via FastMCP client

🔐 Security Pro Tip: Don’t skip authentication. IAM integration is your shield.

Real Example: A Math MCP Server

python

@mcp.tool()
def add(a: int, b: int) -> int:
    return a + b

@mcp.tool()
def subtract(a: int, b: int) -> int:
    return a - b        

✅ Deployed to Cloud Run

✅ Tested via proxy

✅ Fully accessible, remotely authenticated, and secure

Enterprise Takeaways

For Dev Leads & Architects: Centralize tool usage across global teams. Stop reinventing the wheel on every developer machine.

For AI/ML Engineers: Decouple LLM prompting from deterministic logic. Use tools, not tokens.

For IT & Security Teams: Achieve compliance with fine-grained access via IAM and secured endpoints.

Resources & Links

• 🔗 Model Context Protocol Intro
• 🧾 Official Google Guide to Hosting MCP on Cloud Run
• 🧰 FastMCP GitHub

Takeaways

Deploying a remote MCP server is no longer a weekend project. In under 10 minutes, you can make your APIs LLM-ready, contextually aware, and enterprise-scalable.

As AI agents move beyond chat into true autonomy, context is everything. This guide equips you with the foundation to start building serious, contextual AI applications today.

🧠 Let’s stop asking LLMs to guess. Let’s start giving them tools.