Remote MCP with Azure Functions

Prerequisites

• Node.js version 18 or higher
• Azure Functions Core Tools >= 4.0.7030
• Azure Developer CLI
• Visual Studio Code (recommended)
  • Azure Functions extension
  • GitHub Copilot (optional, for testing with Copilot Chat)

Prepare your local environment

1. Clone or navigate to your project directory
2. Install dependencies:

   npm install
   

Available MCP Tools

This MCP server provides three tools for managing project contributions:

• list_projects - Enumerate all projects with contribution counts and team member information
• get_contributions - Filter and retrieve contributions by project name or team member
• add_contribution - Record a new contribution with project, member, summary, and optional attachments

Run your MCP Server locally from the terminal

1. Install dependencies (if not already done)

   npm install
   

2. Build the project

   npm run build
   

3. Start the Functions host locally:

   func start
   

The local MCP endpoint will be available at: http://0.0.0.0:7071/runtime/webhooks/mcp

Note: In Azure, the endpoint includes authentication: /runtime/webhooks/mcp?code=<system_key>

Test the MCP Server locally

Using MCP Inspector

1. In a new terminal window, install and run MCP Inspector

   npx @modelcontextprotocol/inspector node build/index.js
   

2. CTRL+Click the displayed URL (typically http://0.0.0.0:5173/#resources) to open MCP Inspector

3. Set the transport type to http

4. Set the URL to your local MCP endpoint:

   http://0.0.0.0:7071/runtime/webhooks/mcp
   

5. Click Connect and then List Tools to see all available tools

6. Test each tool by clicking it and providing sample parameters:

   • list_projects - No parameters needed, returns all projects
   • get_contributions - Optional: filter by project_name or author_name
   • add_contribution - Required: MemberId, ProjectId, Summary, Description; Optional: AttachmentUrl

When finished, press Ctrl+C in both terminal windows to stop the processes.

Using VS Code GitHub Copilot

1. Start the Functions host locally (as shown above)
2. In VS Code, open the Chat panel with GitHub Copilot
3. Ask Copilot to use the MCP tools:

   List all the projects we have
   

   Show me all contributions from Marcus Thorne
   

   Add a new contribution to the Sorsogon Community Innovation Labs project
   

4. When prompted, click Continue to allow the tool execution
5. Copilot will execute the MCP tool and display the results

Deploy to Azure for Remote MCP

Optionally, enable VNet for additional network isolation before deployment:

azd env set VNET_ENABLED true

Run these azd commands to provision Azure resources and deploy your code:

azd provision

Wait a few minutes for access permissions to take effect, then deploy:

azd deploy

This will:

• Create an Azure Function App with your MCP server
• Set up all required infrastructure (App Service plan, Storage, etc.)
• Deploy your TypeScript code to Azure
• Configure the MCP system key for secure access

Note API Management and App Service built-in authentication can be used for enhanced security and OAuth provider integration.

Get your MCP System Key

After azd deploy completes, you need the MCP system key to authenticate requests.

Option 1: Get from Azure CLI (Recommended)

# Get your function app name from azd
FUNCTION_APP_NAME=$(azd env get-values | grep AZURE_FUNCTION_APP_NAME | cut -d'=' -f2)
RESOURCE_GROUP=$(azd env get-values | grep AZURE_RESOURCE_GROUP | cut -d'=' -f2)

# Retrieve the mcp_extension system key
MCP_KEY=$(az functionapp keys list \
  --resource-group $RESOURCE_GROUP \
  --name $FUNCTION_APP_NAME \
  --query "systemKeys.mcp_extension" -o tsv)

echo "MCP Extension Key: $MCP_KEY"

Option 2: Get from Azure Portal

1. Go to Azure Portal
2. Search for your Function App (name from azd env get-values)
3. Navigate to Functions → App Keys
4. Copy the mcp_extension key from the System Keys section

Connect to your remote MCP server in Azure

After deployment, your MCP server is accessible at:

https://<funcappname>.azurewebsites.net/runtime/webhooks/mcp?code=<your-mcp-extension-system-key>

Example with actual values:

https://mcp-scil-aegccgcze5bsdkez.azurewebsites.net/runtime/webhooks/mcp?code=X9kL2mN5pQ8rT3vW6xY1zA4bC7dE0fG9hI2jK5mL8nO1pR4sT7uV0wX3yZ6aB9cD==

Test in MCP Inspector

Use the key in the URL:

https://<funcappname>.azurewebsites.net/runtime/webhooks/mcp?code=<your-mcp-extension-system-key>

Connect in Azure AI Foundry

1. Go to Azure AI Foundry
2. Create or open your AI project
3. Navigate to Tools → MCP Servers
4. Add your remote MCP server with:
   • URL: https://<funcappname>.azurewebsites.net/runtime/webhooks/mcp
   • Headers: x-functions-key: <your-mcp-extension-system-key>
5. Your AI models can now call the list_projects, get_contributions, and add_contribution tools

Configure Agent with Context

To help your AI agent understand how to use the MCP tools effectively:

1. Create a new Agent in your project

2. In the Configuration panel, set:

   • Display name: "Contribution Manager"
   • Description:

     A contribution tracking assistant that helps manage project contributions 
     across the Sorsogon Community Innovation Labs ecosystem. I can help you:
     - List all available projects and see who's contributing
     - Search for specific contributions by project or team member
     - Record new contributions with details like summary, description, and attachments
     
     Use me to track community engagement, project progress, and team member activities.
     

   • Starter prompts (optional examples):
     1. "List all projects and show me the contribution counts"
     2. "Show me all contributions from Marcus Thorne"
     3. "Add a new contribution to the Sorsogon Community Innovation Labs project"

3. Enable the MCP tools in your agent's Tools configuration

4. Test the agent with your starter prompts

Connect in VS Code GitHub Copilot

Set up in your mcp.json:

{
    "inputs": [
        {
            "type": "promptString",
            "id": "functions-mcp-extension-system-key",
            "description": "Azure Functions MCP Extension System Key",
            "password": true
        },
        {
            "type": "promptString",
            "id": "functionapp-name",
            "description": "Azure Function App Name"
        }
    ],
    "servers": {
        "remote-mcp-function": {
            "type": "http",
            "url": "https://${input:functionapp-name}.azurewebsites.net/runtime/webhooks/mcp",
            "headers": {
                "x-functions-key": "${input:functions-mcp-extension-system-key}"
            }
        }
    }
}

Redeploy your code updates

After making changes to your MCP tools or handlers, rebuild and redeploy:

npm run build
azd deploy

Note: Your function app code is always updated with the latest deployment package. No manual file replacement needed.

Clean up resources

When you're done working with your function app and related resources, you can use this command to delete the function app and its related resources from Azure and avoid incurring any further costs:

azd down

Source Code Structure

The MCP server is organized with Azure Functions' native decorator pattern:

src/
├── index.ts                 # Entry point - imports all tools
├── functions/
│   ├── data/
│   │   └── contributionData.ts    # Shared mock data (projects, contributions, members)
│   └── tools/
│       ├── listProjectsTool.ts    # list_projects MCP tool handler
│       ├── addContributionTool.ts # add_contribution MCP tool handler
│       └── getContributionsTool.ts# get_contributions MCP tool handler

Next Steps

• Enhance Functionality: Add more tools for project management or analytics
• Persist Data: Replace in-memory mock data with a database (Cosmos DB, SQL, etc.)
• Add Security: Enable Azure API Management or App Service authentication
• Enable VNet: Use azd env set VNET_ENABLED true for network isolation
• Monitor & Log: Check Azure Application Insights for tool usage patterns
• Scale: Optimize performance with higher tier App Service plans
• Learn more: Explore MCP documentation and Azure Functions with MCP