MCP Client Configuration with Azure App Configuration

Bridging the Development-Production Gap

Developers predominantly work with MCP servers locally today using tools like Claude Desktop or VSCode. While these environments are excellent for experimentation and development, transitioning MCP-powered applications to production environments presents significant challenges:

• Configuration Management: Local configurations (like VSCode's mcp.json) don't translate seamlessly to production environments
• Scalability Concerns: Production deployments introduce configuration sprawl when MCP settings must be duplicated across every service instance, creating maintenance overhead and inconsistency risks
• Security Requirements: Moving beyond local development necessitates proper authentication and authorization
• Operational Flexibility: The ability to update configurations without redeployment becomes critical

The ConfiguredMcpClientManager Solution

The ConfiguredMcpClientManager implementation addresses these challenges by:

• Centralizing Configuration: Storing all MCP server configurations in Azure App Configuration
• Enabling Dynamic Updates: Allowing changes to MCP servers without application redeployment
• Supporting Environment-Specific Setups: Maintaining separate configurations for development, testing, and production
• Facilitating Deployment Strategies: Enabling blue-green deployments, canary releases, and A/B testing of different model configurations

Key Capabilities

• Configuration Abstraction: Write your application once and connect to different MCP servers based on environment
• Auto-Refresh Mechanism Configurations update automatically at configurable intervals
• Seamless Authentication: Integration with Microsoft Entra ID for secure access to Azure resources
• Deployment Flexibility: Support for various MCP server types (stdio, sse, etc.) with environment-specific parameters

Setup Azure App Configuration

1. Create an Azure App Configuration store in the Azure Portal
2. Add your MCP server configurations as key-value pairs:

Example: Azure MCP Server (stdio type)

Key: MCP:AzureMCPServer

Value:

{
    "name": "Azure MCP Server",
    "type": "stdio",
    "command": "npx",
    "args": [
        "-y",
        "@azure/mcp@latest",
        "server",
        "start"
    ]
}

Example: Remote SSE Server

Key: MCP:Fetch

Value:

{
    "name": "Fetch",
    "type": "sse",
    "url": "https://remote.mcpservers.org/fetch"
}

Implementation Best Practices

1. Environment-Specific Configurations:

   • Use labels in Azure App Configuration to separate dev/test/prod environments
   • Example: MCP:AzureMCPServer;env=production vs MCP:AzureMCPServer;env=development

2. Monitoring and Observability:

   • Add telemetry to track model performance in production
   • Use configuration to enable/disable detailed logging

3. Rollback Strategy:

   • Maintain previous working configurations with version labels
   • Implement quick rollback procedures for production issues

4. Configuration Validation:

   • Validate new configurations before applying them to production
   • Implement circuit breakers to prevent cascading failures