Python MCP Server Template (FastMCP + Docker/Devcontainer)

This is a template project for efficiently developing Python-based Model Context Protocol (MCP) servers using FastMCP and Docker (Devcontainer). It provides samples for basic MCP server functionalities (e.g., an add tool and a greeting resource), utilizes type hints, includes Ruff for static analysis and formatting, a test environment, and supports a ready-to-use development environment setup.

This repository is a Streamable-ready End-to-End (E2E) test template for MCP written in Python and comes with a Docker/Devcontainer setup out of the box.

---

Detailed explanatory articles have been published!

• English (dev.to): Python MCP Remote Server: The Dawn of the Streamable HTTP Era, with a Minimalist Template
• Japanese (Qiita): リモートサーバーで動かす Python MCP Server — Streamable HTTP 時代の始まり ~ uv / Docker / pytest を備えたミニマルテンプレート付き ~

---

Table of Contents

• Features
• Prerequisites
• Setup (Local Environment)
  • Using Ruff (Linter/Formatter) Locally
• Running the Server (Local Environment)
• Running the Client (Local Environment)
• Development with Devcontainer (Recommended)
• Running with Docker
• Running Tests (pytest)
  • Mechanism
  • Execution
• Connecting with MCP Inspector
• Current Features
• License

Features

• MCP Python SDK: Built upon the official Model Context Protocol Python SDK.
• FastMCP Based: Uses FastMCP (part of the SDK), a lightweight and fast MCP server library for Python.
• Docker & Devcontainer: Provides containerization with Docker and a consistent development environment with VS Code Dev Containers.
  • The container runs as a non-root user (appuser) for enhanced security.
• Package Management with uv: Employs uv, a fast Python package installer and resolver.
• Static Analysis and Formatting: Linting and formatting with Ruff are pre-configured.
• Test Environment: Sample asynchronous tests using pytest and pytest-asyncio with an execution environment.
• Type Hints: Actively uses type hints throughout the project to improve code robustness and readability.
• Configurable Logging: Log output can be controlled via the LOG_LEVEL environment variable (e.g., DEBUG, INFO, WARNING).

Prerequisites

• Python (3.10 or higher, see pyproject.toml. Dockerfile uses 3.13-slim)
• uv (Python package management tool)
• Docker (for container execution)
• (Optional) VS Code and Dev Containers extension (for Devcontainer usage)

Setup (Local Environment)

1. Create a new repository from this template or clone the repository. Use the "Use this template" button on GitHub or clone as follows:

   git clone https://github.com/akitana-airtanker/mcp-python-streamable-e2e-test-template.git <your-project-name>
   cd <your-project-name>
   

   Replace <your-project-name> with your project's name.

2. Create a virtual environment and install dependencies. Run the following commands in the project's root directory:

   uv venv
   uv pip install -e ".[test,dev]" # Install test and dev dependencies as well
   

   This creates a virtual environment in the .venv directory and installs the necessary packages (including those for testing and development) based on pyproject.toml.

Using Ruff (Linter/Formatter) Locally

This template uses Ruff. To use it locally:

# Install Ruff (should already be installed with dev dependencies, but can be installed individually)
# uv pip install ruff
# Check code
ruff check .
# Format code
ruff format .

It's recommended to set up pre-commit hooks to automate this:

pre-commit install

Running the Server (Local Environment)

1. Activate the virtual environment (if not already activated).

   source .venv/bin/activate
   

2. Start the server.

   mcp-server-demo
   

   The server will start at http://0.0.0.0:8000 and listen for MCP requests. You can change the log verbosity by setting the LOG_LEVEL environment variable (e.g., LOG_LEVEL=DEBUG mcp-server-demo).

Running the Client (Local Environment)

1. Ensure the server is running.
2. Open another terminal and activate the virtual environment.

   cd <your-project-name> # Navigate to the project root directory
   source .venv/bin/activate
   

3. Run the client.

   mcp-client-demo
   

   The client will connect to the server, call the add tool, and print the result (Result of add(10, 5): 15) to the console. You can control the client's output verbosity using the --quiet or --verbose flags:

   mcp-client-demo --quiet   # Suppresses print statements, logs WARNING and above
   mcp-client-demo --verbose # Enables DEBUG logging and print statements
   

   The LOG_LEVEL environment variable can also be used to set the logging level, with CLI flags taking precedence if specified.

Development with Devcontainer (Recommended)

Using VS Code Dev Containers allows for an easy setup of a consistent development environment with all necessary tools and configurations.

1. Prerequisites:

   • Visual Studio Code
   • Docker Desktop
   • VS Code's Dev Containers extension (recommended, though VS Code may automatically suggest installing it)

2. Open the Development Container:

   • Open this project folder in VS Code.
   • VS Code will detect the .devcontainer/devcontainer.json file and show a notification "Reopen in Container" in the bottom right. Click this notification.
   • Alternatively, open the Command Palette (Ctrl+Shift+P or Cmd+Shift+P), type "Dev Containers: Reopen in Container", and execute it.

3. Working inside the Container:

   • Once the container is built and started, VS Code will automatically connect to the project inside the container.
   • Opening a terminal will open a terminal inside the container.
   • mcp-server-demo and mcp-client-demo can be run directly in this container terminal.

     # Start the server (in the container terminal)
     mcp-server-demo
     
     # Run the client (in another container terminal)
     mcp-client-demo
     

   • Dependencies (including test and dev extras) are installed via the Dockerfile and postCreateCommand in devcontainer.json, so all necessary packages are available when the container starts. Pre-commit hooks are also installed.
   • Linting and formatting with Ruff will be performed automatically due to VS Code settings.
   • The container runs as a non-root user (appuser).
   • Port 8000 is automatically forwarded, so you can access http://localhost:8000/mcp (for MCP Inspector) from your host machine's browser.

Running with Docker

1. Build the Docker image. Run the following command in the project's root directory:

   docker build -t mcp-server .
   

2. Run the Docker container.

   docker run -p 8000:8000 -e LOG_LEVEL=DEBUG mcp-server
   

   This starts the MCP server inside the container and maps it to port 8000 on the host. You can control the server's log level in the container by passing environment variables like -e LOG_LEVEL=DEBUG. The container runs as a non-root user (appuser).

Running Tests (pytest)

This repository includes a minimal test setup located in the tests/ directory (e.g., tests/test_client.py). The tests start the FastMCP server on a dedicated port 8001 (configured in tests/conftest.py) and verify that tool calls work correctly.

Mechanism

• The tests/conftest.py fixture passes MCP_SERVER_PORT=8001 as an environment variable to the child process when starting the server process.
• In src/mcp_python_streamable_e2e_test_template/server.py, the Config class reads MCP_SERVER_PORT and FASTMCP_PORT. If MCP_SERVER_PORT is set and FASTMCP_PORT is not, FASTMCP_PORT (used by FastMCP) defaults to the value of MCP_SERVER_PORT.

  # src/mcp_python_streamable_e2e_test_template/server.py
  # ...
  from .config import Config
  cfg = Config()
  if cfg.mcp_server_port and not cfg.fastmcp_port:
      os.environ.setdefault("FASTMCP_PORT", cfg.mcp_server_port)
  # ...
  

  This ensures that normal startup (mcp-server-demo) defaults to port 8000 (or the value of FASTMCP_PORT / MCP_SERVER_PORT if set), while pytest execution uses port 8001 for the server.

Execution

# Assuming the virtual environment is activated
pytest

If the tests pass, you will see output similar to this:

collected 1 item

test_client.py .                                                       [100%]

============================= 1 passed in X.XXs =============================

---

Connecting with MCP Inspector

With the server running locally or in Docker, you can connect to it using MCP Inspector to verify its behavior.

1. Start MCP Inspector. Run the following command in a new terminal:

   npx --yes @modelcontextprotocol/inspector
   

2. Connect to MCP Inspector. Once MCP Inspector opens in your browser, connect with the following settings:

   • Transport Type: Streamable HTTP
   • URL: http://localhost:8000/mcp

Current Features

• add Tool: Adds two numbers.
  • Example Input: {"a": 10, "b": 5}
  • Example Output: 15 (as TextContent)
• greeting Resource: Returns a greeting for a specified name.
  • Example URI: greeting://World
  • Example Output: "Hello, World!"

---

License

This project is licensed under the MIT License. See the LICENSE file for details.