aagirre92/mcp-SERVER-python-docker-test

MCP Python Docker Test

This project demonstrates a minimal Model Context Protocol (MCP) server in Python, designed to run inside a Docker container and expose simple mathematical tools for use with LLM agents and Microsoft Copilot Studio.

GCP Cloud Run

This repository is connected to GCP Cloud run mcp-server

Main Script

mcp-server-oauth-gcp-cloudRun.py is the main entry point.
It starts an MCP server with OAuth2 authentication using Google Cloud credentials. The script:

• Verifies Google OAuth2 access tokens and restricts access to allowed emails.
• Exposes math tools (add, subtract, multiply, divide, and distance calculations) as MCP tools.
• Supports both primitive-type and Pydantic model arguments (⚠️: only primitive-type tools are visible to Copilot Studio agents. Those tools that have non-primitive types as INPUT are not shown in Copilot Studio Agents).

Quick Start

Build and run with Docker:

docker build -t mcp-python-docker-test .
docker run -p 8080:8080 mcp-python-docker-test

Copilot Studio Integration

• Use a public URL (e.g., via devtunnel) as the host.
• In the Copilot Studio connector definition, set the path to /mcp/ (with the trailing slash). ⚠️ When using FastMCP with from fastmcp import FastMCP I had to put the final slash, but when using it with from mcp.server.fastmcp import FastMCP I did not put! This is very important

Example Swagger:

swagger: '2.0'
info:
  title: GCP-MCP
  version: '1.0'
host: <your-public-url>
basePath: /
schemes:
  - https
paths:
  /mcp/:
    post:
      summary: MCP endpoint
      x-ms-agentic-protocol: mcp-streamable-1.0
      operationId: InvokeMCP
      responses:
        '200':
          description: Success
securityDefinitions: {}

Files

• mcp-server-oauth-gcp-cloudRun.py — Main MCP server with Google OAuth2
• Dockerfile — Containerization instructions
• pyproject.toml — Project dependencies
• server.py — (Alternative) MCP server without