MCP Demo App (Python)

This repository contains a comprehensive demonstration solution for Model Context Protocol (MCP) showcasing all three core MCP capabilities: Tools, Prompts, and Resources. Built with Python and designed for Azure deployment.

Workshop: Step-by-step labs for deploying APIs, MCP servers, and the chat client (Option 1) are in docs/index.md. When published with GitHub Pages, the workshop is available at https://prwani.github.io/mcp_demo_app_python/.

🎯 MCP Capabilities Demonstrated

⚡ Tools (Actions)

• Leave Management: Apply for leave, check balances
• Timesheet Management: Log hours, list entries
• Interactive tool discovery and execution

🎯 Prompts (Templates)

• Leave Prompts: Email templates, policy summaries, calendar planning
• Timesheet Prompts: Submission reminders, project summaries, overtime analysis
• Dynamic prompt generation with user arguments

📚 Resources (Data/Content)

• Leave Resources: Policies, forms, holiday calendars, team status
• Timesheet Resources: Submission guidelines, project codes, templates, utilization reports
• Rich content delivery in multiple formats (text, JSON)

🏗️ Architecture

1. Leave Application

• REST API: FastAPI with SQLAlchemy ORM
• Database: Azure SQL DB (serverless) with leave balances and requests
• MCP Server: Full MCP implementation with tools, prompts, and resources
• Web UI: Interactive leave management interface

2. Timesheet Application

• REST API: FastAPI with timesheet entry management
• Database: Azure SQL DB (serverless) with employee time tracking
• MCP Server: Complete MCP server with timesheet-specific capabilities
• Web UI: Simple timesheet logging interface

3. Chat Client (MCP Client)

• Smart Chat: Azure OpenAI integration with intent detection
• MCP Discovery: Automatic discovery of server capabilities
• Multi-Protocol: Supports tools, prompts, and resources from both servers
• Web UI: Enhanced interface showcasing all MCP features

📁 Project Structure

leave_app/
  api/         # FastAPI REST API for leave management
  mcp_server/  # Complete MCP server (tools, prompts, resources)
  web/         # Interactive web UI for leave operations
  sql/         # Database schema and seed data
timesheet_app/
  api/         # FastAPI REST API for timesheet management  
  mcp_server/  # Full MCP server implementation
  web/         # Timesheet logging web interface
  sql/         # Database schema and seed data
chat_client/
  api/         # Enhanced chat client with MCP discovery
  web/         # Rich web UI showcasing all MCP capabilities
requirements.txt # Python dependencies
.dockerignore    # Docker build optimization
*/Dockerfile.*  # Containerization for each service

🚀 Quick Start

Local Development

# 1. Setup Python environment
python -m venv .venv
source .venv/bin/activate  # Linux/Mac
pip install -r requirements.txt

# 2. Run services locally (separate terminals)
uvicorn leave_app.api.main:app --port 8001 --reload
uvicorn leave_app.mcp_server.server:app --port 8011 --reload
uvicorn timesheet_app.api.main:app --port 8002 --reload  
uvicorn timesheet_app.mcp_server.server:app --port 8012 --reload
uvicorn chat_client.api.main:app --port 8000 --reload

# 3. Open http://localhost:8000 for the enhanced chat interface

Try MCP Features

• Tools: "Apply leave 2025-08-20 to 2025-08-22" or "Log 8 hours on 2025-08-20"
• Prompts: "Generate leave request email" or "Create timesheet reminder"
• Resources: "Show leave policy" or "Get project codes list"
• Discovery: Click "Discover MCP Capabilities" to see all available features

🔧 MCP Endpoints

Tools (Actions)

• POST /mcp/tools/apply_leave - Submit leave requests
• GET /mcp/tools/get_balance - Check leave balances
• POST /mcp/tools/add_timesheet_entry - Log work hours
• GET /mcp/tools/list_timesheet_entries - Retrieve timesheet data

Prompts (Templates)

• GET /mcp/prompts/list - Discover available prompts
• POST /mcp/prompts/get - Generate customized content
• Examples: leave_request_email, timesheet_reminder, policy_summary

Resources (Content)

• GET /mcp/resources/list - Browse available resources
• POST /mcp/resources/read - Access specific content
• Examples: leave policies, project codes, best practices, reports

🌐 Azure Deployment

• Containers: Docker images built with Azure Container Registry
• Hosting: Azure Web Apps for Containers with Linux
• Databases: Azure SQL DB (serverless) with automatic scaling
• AI: Azure OpenAI for intelligent chat interactions
• Security: Managed Identity with role-based access control

🎛️ Configuration

Set these environment variables for full functionality:

# Database connections
LEAVE_DATABASE_URL="azure-sql-connection-string"
TIMESHEET_DATABASE_URL="azure-sql-connection-string"

# MCP server URLs  
LEAVE_MCP_URL="https://leave-mcp.azurewebsites.net"
TIMESHEET_MCP_URL="https://timesheet-mcp.azurewebsites.net"

# Azure OpenAI (optional for enhanced chat)
AZURE_OPENAI_ENDPOINT="https://your-aoai.openai.azure.com/"
AZURE_OPENAI_KEY="your-api-key"
AZURE_OPENAI_DEPLOYMENT="gpt-4o-mini"

💡 What Makes This Special

• Complete MCP Implementation: First demo showcasing tools, prompts, AND resources
• Real-World Use Cases: Practical workplace scenarios with actual business logic
• Intelligent Chat: Azure OpenAI integration with intent detection and fallback heuristics
• Auto-Discovery: Dynamic MCP capability discovery and presentation
• Production Ready: Containerized with Azure deployment, security, and monitoring

License

MIT

For detailed Azure deployment steps, see Azure_setup.md.

📘 Workshop

• The full MCP Workshop (Labs 1–4, concepts-first then commands) is available at docs/index.md and on GitHub Pages when enabled.
  • Quick link in this repo: docs/index.md
  • GitHub Pages URL (after enabling Pages for /docs): https://.github.io//