MCP Git Commit Generator
Generate conventional commit messages from your staged git changes using Model Context Protocol (MCP).
✨ Features
- 🤖 Automatic commit message generation based on staged git diffs
- 📝 Conventional Commits support with auto-detection of type and scope
- 🔄 Multiple transport options - stdio (default) and SSE for different use cases
- 🔍 Inspector UI for interactive testing and debugging
- 🐳 Docker support with pre-built images for easy deployment
- ⚡ Cross-platform - works on macOS, Linux, and Windows
📦 Requirements
- For Docker usage: Docker (for running the server in a container)
- For PyPI/uvx usage: Python >= 3.13.5 and uv (recommended) or pip
- Git (for version control)
- An MCP-compatible client (VS Code with MCP extension, Claude Desktop, Cursor, Windsurf, etc.)
🚀 Installation
You can install and use the MCP Git Commit Generator in multiple ways:
Option 1: Using uvx (Recommended)
The easiest way to use the package is with uvx, which automatically manages the virtual environment:
uvx mcp-git-commit-generator
Option 2: Install from PyPI
pip install mcp-git-commit-generator
Or with uv:
uv pip install mcp-git-commit-generator
Option 3: Using Docker
Use the pre-built Docker image from GitHub Container Registry (no installation required):
docker run -i --rm --mount type=bind,src=${HOME},dst=${HOME} ghcr.io/theoklitosbam7/mcp-git-commit-generator:latest
🛠️ Available Tools
This MCP server provides the following tools to help you generate conventional commit messages:
generate_commit_message
Generates a conventional commit message based on your staged git changes.
Parameters:
repo_path(string, optional): Path to the git repository. If omitted, uses the current directory.commit_type(string, optional): Conventional commit type (e.g.,feat,fix,docs,style,refactor,perf,build,ci,test,chore,revert). If omitted, the type will be auto-detected.scope(string, optional): Scope of the change (e.g., file or module name). If omitted, the scope will be auto-detected based on changed files.
Usage:
- Stage your changes:
git add <files> - Use the tool through your MCP client to generate a commit message
- The tool will analyze your staged changes and generate an appropriate conventional commit message
check_git_status
Checks the current git repository status, including staged, unstaged, and untracked files.
Parameters:
repo_path(string, optional): Path to the git repository. If omitted, uses the current directory.
Usage:
Use this tool to get an overview of your current git repository state before generating commit messages.
🧩 MCP Client Configuration
Configure the MCP Git Commit Generator in your favorite MCP client. You have multiple options:
- Using uvx (recommended - automatically manages dependencies)
- Using Docker (no local Python installation required)
- Using local Python installation (for development)
VS Code
Add one of the following configurations to your VS Code mcp.json file (usually located at .vscode/mcp.json in your workspace):
Using uvx (Recommended)
{
"servers": {
"mcp-git-commit-generator": {
"command": "uvx",
"args": ["mcp-git-commit-generator"]
}
}
}
Using Docker
{
"servers": {
"mcp-git-commit-generator": {
"command": "docker",
"args": [
"run",
"-i",
"--rm",
"--mount",
"type=bind,src=${userHome},dst=${userHome}",
"ghcr.io/theoklitosbam7/mcp-git-commit-generator:latest"
]
}
}
}
If you want to put the configuration in your user settings.json file, you can do so by adding:
{
"mcp": {
"servers": {
"mcp-git-commit-generator": {
"command": "uvx",
"args": ["mcp-git-commit-generator"]
}
}
}
}
Cursor
Add one of the following to your Cursor MCP configuration file (usually located at ~/.cursor/mcp.json):
Cursor with uvx (Recommended)
{
"mcpServers": {
"mcp-git-commit-generator": {
"command": "uvx",
"args": ["mcp-git-commit-generator"]
}
}
}
Cursor with Docker
{
"mcpServers": {
"mcp-git-commit-generator": {
"command": "docker",
"args": [
"run",
"-i",
"--rm",
"--mount",
"type=bind,src=${userHome},dst=${userHome}",
"ghcr.io/theoklitosbam7/mcp-git-commit-generator:latest"
]
}
}
}
Windsurf
Configure Windsurf with one of the following MCP server settings (usually located at ~/.codeium/windsurf/mcp_config.json):
Windsurf with uvx (Recommended)
{
"mcpServers": {
"mcp-git-commit-generator": {
"command": "uvx",
"args": ["mcp-git-commit-generator"]
}
}
}
Windsurf with Docker
{
"mcpServers": {
"mcp-git-commit-generator": {
"command": "docker",
"args": [
"run",
"-i",
"--rm",
"--mount",
"type=bind,src=${userHome},dst=${userHome}",
"ghcr.io/theoklitosbam7/mcp-git-commit-generator:latest"
]
}
}
}
Claude Desktop
Add one of the following to your Claude Desktop configuration file (usually located at
~/Library/Application Support/Claude/claude_desktop_config.json on macOS):
Claude Desktop with uvx (Recommended)
{
"mcpServers": {
"mcp-git-commit-generator": {
"command": "uvx",
"args": ["mcp-git-commit-generator"]
}
}
}
Claude Desktop with Docker
{
"mcpServers": {
"mcp-git-commit-generator": {
"command": "docker",
"args": [
"run",
"-i",
"--rm",
"--mount",
"type=bind,src=${userHome},dst=${userHome}",
"ghcr.io/theoklitosbam7/mcp-git-commit-generator:latest"
]
}
}
}
Note: The
--mountoption in Docker configurations allows the Docker container to access your home directory, enabling it to work with git repositories located anywhere in your file system. When using uvx or pip installations, this mounting is not needed as the tool runs directly on your system. Adjust the mount path if your repositories are located elsewhere when using Docker.
🚀 Quick Start Guide
-
Install the package using one of the methods above:
- Recommended:
uvx mcp-git-commit-generator(or configure in your MCP client) - Alternative:
pip install mcp-git-commit-generator - Docker: Use the configurations above with Docker
- Recommended:
-
Configure your MCP client using one of the configurations above
-
Stage some changes in a git repository:
git add <files> -
Use the tools through your MCP client:
- Use
check_git_statusto see your current repository state - Use
generate_commit_messageto create a conventional commit message
- Use
-
Commit your changes with the generated message
👨💻 Developer Guidelines
The following sections are intended for developers who want to contribute to or modify the MCP Git Commit Generator.
Local Development Setup 🛠️
If you prefer not to use Docker for development, you can run the server locally:
Requirements:
- Python >= 3.13.5
- uv (recommended for dependency management and local development)
- Node.js (for Inspector UI, optional)
- Python Debugger Extension (for debugging, optional)
Note: The MCP CLI dependency is automatically installed via
uv.lockwhen usinguv sync.
Installation:
-
Clone the repository:
git clone https://github.com/theoklitosBam7/mcp-git-commit-generator.git cd mcp-git-commit-generator -
Prepare environment:
There are two approaches to set up the environment for this project. You can choose either one based on your preference.
Note: Reload VSCode or terminal to ensure the virtual environment python is used after creating the virtual environment.
Approach Steps Using uv(Recommended)1. Create virtual environment and install dependencies: uv sync --group dev
2. Run VSCode Command "Python: Select Interpreter" and select the python from.venvdirectory
3. The project is installed in editable mode automatically byuv sync.Using pip1. Create virtual environment: python -m venv .venv
2. Run VSCode Command "Python: Select Interpreter" and select the python from created virtual environment
3. Install dependencies:pip install -e ..
4. Install pip dev dependencies:pip install -r requirements-dev.txt. -
(Optional) Install Inspector dependencies:
cd inspector npm install
📦 Publishing to PyPI
The project includes an automated PyPI publishing workflow (.github/workflows/pypi-publish.yml) that:
- Triggers on: Tag pushes matching
v*.*.*pattern, manual workflow dispatch, or pull requests to main - Builds: Python package distributions using the
buildpackage - Publishes: Automatically publishes to PyPI using trusted publishing (OIDC) when tags are pushed
To publish a new version:
- Update the version in
pyproject.toml - Create and push a git tag:
git tag vX.Y.Z && git push origin vX.Y.Z - The workflow will automatically build and publish to PyPI
🐳 Building and Running with Docker
You can build and run the MCP Git Commit Generator using Docker. The provided Dockerfile uses a multi-stage build
with uv for dependency management and runs the server as a non-root user for security.
Build the Docker image
docker build -t mcp-git-commit-generator .
Run the server in a container (default: stdio transport)
You can run the published image directly from GitHub Container Registry.
docker run -d \
--name mcp-git-commit-generator \
ghcr.io/theoklitosbam7/mcp-git-commit-generator:latest
By default, the container runs:
mcp-git-commit-generator --transport stdio
If you want to use SSE transport (for Inspector UI or remote access), override the entrypoint or run manually:
docker run -d \
--name mcp-git-commit-generator \
-p 3001:3001 \
--entrypoint mcp-git-commit-generator \
ghcr.io/theoklitosbam7/mcp-git-commit-generator:latest --transport sse --host 0.0.0.0 --port 3001
The server will be available at http://localhost:3001 when using SSE.
🖥️ Running the Server Locally
To run locally (without Docker):
- Set up your uv or Python environment as described in the Local Development Setup section.
- From the project root, run:
Using uv (Recommended)
# Run with uv (uses uv.lock for consistent dependencies)
uv run mcp-git-commit-generator
# Or with SSE transport
uv run mcp-git-commit-generator --transport sse
Using uv run with module
uv run -m mcp_git_commit_generator --transport sse
mcp-git-commit-generator (installed in environment)
# Default stdio transport
mcp-git-commit-generator
# With SSE transport
mcp-git-commit-generator --transport sse
Using Python directly
python -m mcp_git_commit_generator --transport sse
You can specify other options, for example:
python -m mcp_git_commit_generator --transport sse --host 0.0.0.0 --port 3001 -v
The server listens on
0.0.0.0:3001by default when using SSE, or as specified by the options above.
Note:
- If you want to use the CLI entrypoint, ensure the package is installed and your environment is activated.
- Do not use positional arguments (e.g.,
python -m mcp_git_commit_generator sse); always use options like--transport sse. - Available arguments with their values are:
--transport: Transport type (e.g.,stdio(default),sse).--host: Host to bind the server (default:0.0.0.0).--port: Port to bind the server (default:3001).-v,--verbose: Verbosity level (e.g.,-v,-vv).
🔎 Start the Inspector UI
From the inspector directory:
npm run dev:inspector
The Inspector UI will be available at
http://localhost:5173.
🧪 Running Tests
The project includes comprehensive unit tests to ensure reliability:
# Run all tests (recommended when using uv)
uv run pytest
# Or using pytest directly
pytest
# Run tests with verbose output
uv run pytest -v
# Run tests with coverage
uv run pytest --cov=src/mcp_git_commit_generator
# Run specific test file
uv run pytest tests/test_server.py
Test Coverage:
- ✅ Tool validation with invalid repository paths
- ✅ Staged and unstaged change detection
- ✅ Git status reporting
- ✅ Commit message generation workflows
- ✅ Error handling for git command failures
🗂️ Project Structure
.
├── .github/ # GitHub workflows and issue templates
├── .gitignore
├── .markdownlint.jsonc
├── .python-version
├── .vscode/ # VSCode configuration
├── LICENSE
├── README.md
├── pyproject.toml # Python project configuration
├── requirements-dev.txt # Development dependencies (for pip users)
├── uv.lock # Dependency lock file for reproducible builds (used by uv sync)
├── Dockerfile # Docker build file
├── build/ # Build artifacts
├── src/ # Python source code
│ └── mcp_git_commit_generator/
│ ├── __init__.py # Main entry point
│ ├── __main__.py # CLI entry point
│ └── server.py # Main server implementation
└── inspector/ # Inspector related files
├── package.json # Node.js dependencies
└── package-lock.json
⚙️ Advanced MCP Server Configuration for Development
The .vscode/mcp.json file configures how VS Code and related tools connect to your MCP Git Commit Generator server.
This file defines available server transports and their connection details, making it easy to switch between
different modes (stdio is default, SSE is optional) for development and debugging.
Example Development mcp.json
{
"servers": {
"mcp-git-commit-generator": {
"command": "docker",
"args": [
"run",
"-i",
"--rm",
"--mount",
"type=bind,src=${userHome},dst=${userHome}",
"ghcr.io/theoklitosbam7/mcp-git-commit-generator:latest"
]
},
"sse-mcp-git-commit-generator": {
"type": "sse",
"url": "http://localhost:3001/sse"
},
"stdio-mcp-git-commit-generator": {
"type": "stdio",
"command": "${command:python.interpreterPath}",
"args": ["-m", "mcp_git_commit_generator", "--transport", "stdio"]
},
"uvx-mcp-git-commit-generator": {
"command": "uvx",
"args": ["mcp-git-commit-generator"]
}
}
}
- mcp-git-commit-generator: Runs the server in a Docker container (default: stdio transport), using the published image.
- sse-mcp-git-commit-generator: Connects to the MCP server using Server-Sent Events (SSE) at
http://localhost:3001/sse. Only useful if you run the server with--transport sse. - stdio-mcp-git-commit-generator: Connects using standard input/output (stdio), running the server as a subprocess. This is the default and recommended for local development and debugging.
- uvx-mcp-git-commit-generator: Uses uvx to automatically install and run the package from PyPI.
🐞 Debugging the MCP Server
Notes:
- MCP Inspector is a visual developer tool for testing and debugging MCP servers.
- All debugging modes support breakpoints, so you can add breakpoints to the tool implementation code.
- You can test tool arguments directly in the Inspector UI: When using the Inspector, select a tool and provide arguments in the input fields to simulate real usage and debug argument handling.
| Debug Mode | Description | Steps to debug |
|---|---|---|
| MCP Inspector | Debug the MCP server using the MCP Inspector. | 1. Install Node.js 2. Set up Inspector: cd inspector && npm install 3. Open VS Code Debug panel. Select Debug in Inspector (Edge) or Debug in Inspector (Chrome). Press F5 to start debugging.4. When MCP Inspector launches in the browser, click the Connect button to connect this MCP server.5. Then you can List Tools, select a tool, input parameters (see arguments above), and Run Tool to debug your server code. |
⚙️ Default Ports and Customizations
| Debug Mode | Ports | Definitions | Customizations | Note |
|---|---|---|---|---|
| MCP Inspector | 3001 (Server, SSE only); 5173 and 3000 (Inspector) | tasks.json | Edit launch.json, tasks.json, __init__.py, mcp.json to change above ports. | N/A |
💬 Feedback
If you have any feedback or suggestions, please open an issue on the MCP Git Commit Generator GitHub repository
📖 Troubleshooting
Common Issues
- "Path is not a valid git repository": Ensure you're in a directory with a
.gitfolder - "No staged changes found": Run
git add <files>to stage your changes first - "Git is not installed": Install Git from git-scm.com
- Docker permission issues: Ensure Docker can access your home directory
- MCP connection fails: Verify your client configuration matches the examples above
Getting Help
- Check the Issues page for solutions
- Use the Inspector UI for interactive debugging
- Run
uv run pytest -vto verify your installation (recommended) - Or run
pytest -vif using pip
📄 License
MIT License © 2025 Theoklitos Bampouris