simple_mcp_server

Vinipallakki/simple_mcp_server

3.1

If you are the rightful owner of simple_mcp_server and would like to certify it and/or have it hosted online, please leave a comment on the right or send an email to dayong@mcphub.com.

This project demonstrates how to build and deploy a secure Model Context Protocol (MCP) server using FastMCP on Google Cloud Run, with tools for querying zoo animals.

Tools
2
Resources
0
Prompts
0

Zoo MCP Server on Cloud Run

This project shows how to build and deploy a secure Model Context Protocol (MCP) server using FastMCP on Google Cloud Run. It includes tools for querying zoo animals by species and getting details for individual animals. The server is set up to require authentication so only authorized clients can invoke its tools.

🚀 What you'll learn

  • How to build a simple MCP server with FastMCP
  • How to containerize it and deploy to Cloud Run
  • Securing the service endpoint so that only authenticated requests are allowed
  • Connecting a client (e.g. Gemini CLI) to your remote MCP server

📋 Project structure


.
├── Dockerfile
├── pyproject.toml
├── server.py
└── README.md
  • server.py — Python MCP server code with tools for animal lookup
  • Dockerfile — To containerize the application
  • pyproject.toml — Dependency and project config
  • README.md — This file

🔧 Prerequisites

  • A Google Cloud account
  • A Google Cloud project with billing enabled
  • gcloud SDK installed and configured
  • (Optional) Gemini CLI if you want to test from CLI

⚙️ Setup Guide

Below are the high-level steps to setup and deploy the MCP server.

1. Enable required APIs

gcloud services enable \
  run.googleapis.com \
  artifactregistry.googleapis.com \
  cloudbuild.googleapis.com

2. Initialize the Python project

mkdir zoo-mcp-server && cd zoo-mcp-server

# Create project config (assuming using `uv` for project scaffolding)
uv init --description "Zoo MCP server on Cloud Run" --bare --python 3.13

Add fastmcp as a dependency. For example in pyproject.toml:

[project]
name = "zoo-mcp-server"
version = "0.1.0"
description = "Zoo MCP Server example"
requires-python = ">=3.13"

[dependencies]
fastmcp = "2.11.1"

3. Write the server code (server.py)

Here’s a simplified example:

import asyncio
import logging
import os
from typing import List, Dict, Any

from fastmcp import FastMCP

logger = logging.getLogger(__name__)
logging.basicConfig(
    format="[%(levelname)s]: %(message)s",
    level=logging.INFO
)

mcp = FastMCP("Zoo Animal MCP Server")

# In-memory data store
ZOO_ANIMALS = [
    {"species": "lion", "name": "Leo", "age": 7, "enclosure": "The Big Cat Plains", "trail": "Savannah Heights"},
    {"species": "penguin", "name": "Waddles", "age": 2, "enclosure": "The Arctic Exhibit", "trail": "Polar Path"},
    # … add more animals …
]

@mcp.tool()
def get_animals_by_species(species: str) -> List[Dict[str, Any]]:
    logger.info(f">>> Tool: 'get_animals_by_species' called for '{species}'")
    return [animal for animal in ZOO_ANIMALS if animal["species"].lower() == species.lower()]

@mcp.tool()
def get_animal_details(name: str) -> Dict[str, Any]:
    logger.info(f">>> Tool: 'get_animal_details' called for '{name}'")
    for animal in ZOO_ANIMALS:
        if animal["name"].lower() == name.lower():
            return animal
    return {}

if __name__ == "__main__":
    port = int(os.getenv("PORT", 8080))
    logger.info(f"Starting MCP server on port {port}")
    asyncio.run(
        mcp.run_async(
            transport="http",
            host="0.0.0.0",
            port=port,
        )
    )

4. Dockerize

Create Dockerfile:

# Use official Python image
FROM python:3.13-slim

# Copy uv tool (if using uv scaffolding)
COPY --from=ghcr.io/astral-sh/uv:latest /uv /uvx /bin/

# Copy project
COPY . /app
WORKDIR /app

# Ensure logs appear promptly
ENV PYTHONUNBUFFERED=1

# Install dependencies
RUN uv sync

# Expose port
EXPOSE $PORT

# Command to run the server
CMD ["uv", "run", "server.py"]

5. Deploy to Cloud Run

gcloud run deploy zoo-mcp-server \
  --no-allow-unauthenticated \
  --region=<YOUR_REGION> \
  --source=. \
  --labels=env=prod,project=zoo-mcp
  • The flag --no-allow-unauthenticated ensures that only authenticated requests can invoke the MCP server.
  • Make sure to choose your preferred region.

🔐 Setting up access & client

Once deployed, you’ll want to allow your user (or service account) to invoke the service:

gcloud projects add-iam-policy-binding $GOOGLE_CLOUD_PROJECT \
  --member=user:$(gcloud config get-value account) \
  --role='roles/run.invoker'

You can also retrieve an ID_TOKEN for use with clients:

export ID_TOKEN=$(gcloud auth print-identity-token)

Configuring Gemini CLI (or other MCP clients)

In your client config (e.g. ~/.gemini/settings.json), add:

{
  "mcpServers": {
    "zoo-remote": {
      "httpUrl": "https://<YOUR_CLOUD_RUN_SERVICE_URL>/mcp/",
      "headers": {
        "Authorization": "Bearer <YOUR_ID_TOKEN>"
      }
    }
  },
  "selectedAuthType": "cloud-shell",
  "hasSeenIdeIntegrationNudge": true
}

🧰 Logs & Debugging

  • To see if tool calls are working:

    gcloud run services logs read zoo-mcp-server --region=<YOUR_REGION> --limit=10

  • A common issue: authentication errors — make sure your client’s bearer token is valid. If the ID token times out, re-fetch it.

🧼 Cleanup

If you want to avoid ongoing charges:

gcloud projects delete $GOOGLE_CLOUD_PROJECT

Or manually delete:

  • The Cloud Run service
  • Artifact Registry repository storing your container images

📌 References

Happy building! 🎉 If you need help adjusting for your own use-case (different tools, data source, region, etc.), I can help you tweak this.