dev-kit

The Dev-Kit MCP Server provides a secure, scoped environment for performing common development operations inside a project root. It exposes tools for file manipulation, Makefile execution, and Git operations, all within an authorized directory to prevent unsafe file access. This makes it a solid backend for AI-assisted tooling (e.g., VS Code Copilot) by offering a defined set of MCP-compatible actions that clients can call in a controlled way. You can pair it with a client that lists available tools, invokes them with parameters, and receives structured results about file changes, command outputs, and Git state.

To use the server, install the package and run the server binary/module. The server accepts a root directory and optional predefined command sets via a TOML file. You can then query the list of tools, call specific tools (like create_dir, edit_file, git_status, or exec_make_target), and receive results that describe success, changes made, and any console output. This setup is particularly helpful when integrating with AI-assisted editors or automation pipelines that need to operate within a project workspace without broad system access.

Prerequisites:

• Python 3.8+ (or a compatible Python environment)
• pip (Python package installer)

Install the MCP server package:

pip install dev-kit-mcp-server

Run the server (examples):

# Run the MCP server from the installed package
python -m dev_kit_mcp_server.mcp_server --root-dir=workdir

# Optional: use a custom TOML file for predefined commands
python -m dev_kit_mcp_server.mcp_server --root-dir=workdir --commands-toml=custom_commands.toml

If you prefer an alternative invocation (and uv/virtual environments are in use):

uv run python -m dev_kit_mcp_server.mcp_server --root-dir=workdir

Notes:

• The server enforces operations within the specified root directory to ensure security. Always provide a valid root-dir that your clients should operate within.
• You can define predefined commands in a TOML file (default is pyproject.toml under [tool.dkmcp.commands]); this lets you reuse common workflows like tests, linting, or builds.
• For large repositories or long-running operations, consider isolating commands and limiting output length to avoid excessive responses.
• If you encounter issues with Python environments, ensure your PATH includes the Python interpreter and that the correct virtual environment is active when launching the server.
• When integrating with MCP clients, ensure the client handles tool listing, parameter validation, and error reporting to provide robust UX.