Octo Server

A Python MCP server hosted with the Azure Functions MCP extension. The app exposes National Weather Service tools through Azure Functions MCP tool triggers.

This repository follows the Azure Functions MCP extension model, not the self-hosted custom-handler model.

Tools

Tool	Description
echo	Returns a message unchanged
get_weather_forecast	Gets a NWS forecast for a US latitude and longitude
get_current_conditions	Gets current NWS observations for a US latitude and longitude
get_weather_alerts	Gets active NWS alerts for a two-letter US state

The weather tools use the public National Weather Service API. No API key is required, but locations must be in the United States.

Project Structure

function_app.py              # Azure Functions MCP tool triggers
host.json                    # Azure Functions host config with experimental MCP extension bundle
local.settings.example.json  # Local settings template; copy to local.settings.json
requirements.txt             # Azure deployment dependencies
pyproject.toml               # Local uv project metadata
.funcignore                  # Files excluded from deployment packages
README.md

The important files for this deployment model are:

• function_app.py
• host.json
• local.settings.json
• requirements.txt
• .funcignore

Azure Setup

Create the Function App in the Azure Portal.

Use these settings:

Setting	Value
Hosting plan	Flex Consumption
Operating system	Linux
Instance memory	512 MB is fine for this sample
Runtime stack	Python
Python version	3.11 or 3.12
Storage	Create or select an Azure Storage account
Application Insights	Recommended
Authentication level	Function

After creating the app, confirm the Overview page shows:

• A domain like <your-function-name>.azurewebsites.net
• Linux as the operating system
• Flex Consumption as the plan

Local Setup

Install prerequisites:

• Python 3.11 or 3.12
• Azure Functions Core Tools v4
• Azure CLI
• uv
• Node.js
• Podman or Docker

Copy local settings:

Copy-Item local.settings.example.json local.settings.json

Create the Python environment:

uv venv
uv pip install -r requirements.txt

If you prefer to add packages through uv:

uv add azure-functions httpx

Run Local Storage

Azure Functions needs storage even when running locally. This example uses Azurite through Podman.

Start the Podman machine:

podman machine start

Run Azurite:

podman run -d --name azurite `
  -p 10000:10000 `
  -p 10001:10001 `
  -p 10002:10002 `
  mcr.microsoft.com/azure-storage/azurite

Check the container:

podman ps -a

If the container already exists but is stopped:

podman start azurite

Run Locally

Start the function host:

func start

Expected output includes a line similar to:

MCP server SSE endpoint: http://localhost:7071/runtime/webhooks/mcp/sse

The local MCP endpoint is:

http://localhost:7071/runtime/webhooks/mcp/sse

Test Locally

Start MCP Inspector:

npx @modelcontextprotocol/inspector

Open the Inspector URL that prints in the terminal, usually:

http://127.0.0.1:6274

Connect to:

http://localhost:7071/runtime/webhooks/mcp/sse

Click List Tools. You should see the four tools listed above.

Deploy To Azure

Sign in:

az login

Set the Function App values:

$FUNCTION_APP_NAME = "<your-function-name>"
$RESOURCE_GROUP = "<your-resource-group>"
$ZIP_FILE = "octo-mcp.zip"

Create the deployment zip:

Remove-Item $ZIP_FILE -Force -ErrorAction SilentlyContinue

$exclude = @(
  ".venv",
  ".git",
  ".vscode",
  ".azurite",
  "local.settings.json",
  "__pycache__",
  $ZIP_FILE
)

Get-ChildItem -Force |
  Where-Object { $exclude -notcontains $_.Name } |
  Compress-Archive -DestinationPath $ZIP_FILE -Force

Confirm the zip contains the required files:

tar -tf $ZIP_FILE | Select-String "function_app.py|host.json|requirements.txt|.funcignore"

Deploy the zip with remote build:

az functionapp deployment source config-zip `
  --src $ZIP_FILE `
  --name $FUNCTION_APP_NAME `
  --resource-group $RESOURCE_GROUP `
  --build-remote true

After deployment, check the Function App in the Azure Portal. Under Functions, you should see:

• echo
• get_weather_forecast
• get_current_conditions
• get_weather_alerts

Test In Azure

In the Azure Portal, open the Function App.

Go to:

Functions > App keys

Under System keys, copy the key named:

mcp_extension

The cloud MCP endpoint is:

https://<your-function-name>.azurewebsites.net/runtime/webhooks/mcp/sse?code=<mcp_extension_key>

Start MCP Inspector:

npx @modelcontextprotocol/inspector

Connect to the cloud endpoint above and click List Tools.

Useful PowerShell Commands

View Function App deployment logs:

az webapp log tail `
  --name $FUNCTION_APP_NAME `
  --resource-group $RESOURCE_GROUP

List app settings:

az functionapp config appsettings list `
  --name $FUNCTION_APP_NAME `
  --resource-group $RESOURCE_GROUP `
  --output table

Restart the Function App:

az functionapp restart `
  --name $FUNCTION_APP_NAME `
  --resource-group $RESOURCE_GROUP

Troubleshooting

Symptom	Fix
func: command not found	Install Azure Functions Core Tools v4
Local storage error	Start Azurite with Podman and confirm AzureWebJobsStorage is UseDevelopmentStorage=true
Tools do not appear locally	Confirm func start shows the MCP SSE endpoint
Tools do not appear in Azure	Confirm deployment succeeded and the Functions list shows the tool functions
Cloud Inspector cannot connect	Confirm the URL includes /runtime/webhooks/mcp/sse?code=<mcp_extension_key>
mcp_extension key missing	Confirm host.json uses Microsoft.Azure.Functions.ExtensionBundle.Experimental

Notes

• Do not set FUNCTIONS_WORKER_RUNTIME in Azure app settings for Flex Consumption through CLI commands. The portal/runtime stack config handles the cloud runtime. Keep it in local.settings.json for local development.
• Do not use the old /mcp endpoint for this repository. The MCP extension endpoint is /runtime/webhooks/mcp/sse.
• Do not deploy local.settings.json.