octo-mcp-server

A Model Context Protocol (MCP) server built with Python and FastMCP, hosted on Azure Functions using the custom handler pattern. It uses the streamable-http transport and is designed for stateless, serverless deployment on the Flex Consumption plan.

Project structure

server.py              # FastMCP server — all tools defined here
host.json              # Azure Functions custom handler config (required)
pyproject.toml         # Python project metadata and dependencies (uv)
azure.yaml             # Azure Developer CLI (azd) project config
.env.example           # Template for local.settings.json
Dockerfile             # Container image (optional; azd deploy preferred)
infra/                 # Bicep IaC for provisioning Azure resources (optional)

Adding a new tool

Add a decorated function to server.py:

@mcp.tool()
async def my_tool(input: str) -> str:
    """
    Description shown to the MCP client.

    Args:
        input: Description of the parameter.
    """
    result = await call_some_api(input)
    return str(result)

That's it — FastMCP registers the tool automatically.

Local development

Prerequisites

• Python ≥ 3.11

• uv — curl -LsSf https://astral.sh/uv/install.sh | sh

• Azure Functions Core Tools

• (Optional) Azurite for local storage emulation

Setup

1. Create local.settings.json from the example:

cp .env.example local.settings.json

local.settings.json is gitignored and never committed.

2. Start the server:

uv run func start

The MCP server starts on http://localhost:7071. The Azure Functions host proxies all requests to the FastMCP process on port 8000.

Test with MCP Inspector

npx @modelcontextprotocol/inspector@latest http://localhost:7071/mcp

Configure VS Code Copilot

Add to .vscode/mcp.json:

{
  "servers": {
    "local-mcp-server": {
      "type": "http",
      "url": "http://localhost:7071/mcp"
    }
  }
}

Deploy to Azure

Prerequisites

• Azure Developer CLI (azd)

• An Azure subscription

Deploy

azd auth login
azd up

azd up provisions all infrastructure and deploys the Function App code. For code-only updates:

azd deploy

Set environment variables in Azure

az functionapp config appsettings set \
  --name <function-app-name> \
  --resource-group <resource-group> \
  --settings MY_API_KEY=your_value

MCP endpoint URLs

Docker

The Dockerfile is provided for cases where a container image is needed. For standard Azure Functions deployment, use azd up instead.

docker build -t octo-mcp-server .
docker run -p 8000:8000 octo-mcp-server

Environment variables

Add tool-specific secrets (API keys, connection strings) to local.settings.json under Values for local dev, and as Azure App Settings for production.

Architecture notes

• Azure Functions acts as a managed host — it proxies all HTTP traffic to the FastMCP process running on port 8000.

• The configurationProfile: "mcp-custom-handler" in host.json configures route proxying, removes the /api prefix, and sets auth to anonymous (auth is handled by EasyAuth at the platform layer).

• A 404 on the root / at startup is expected — the Functions host pings / but FastMCP doesn't implement it.

• Red log output from the MCP SDK on startup is cosmetic — it writes to stderr, which Functions renders in red.

• stateless_http=True is required for compatibility with the Flex Consumption plan's scale-out behavior.