MCP Code Checker

What is MCP Code Checker?

MCP Code Checker is a Model Context Protocol server that enables AI assistants to run code quality checks on Python projects. It provides a focused set of tools – pylint, pytest, and mypy – scoped to a specified project directory, with structured output and actionable prompts to reduce context load.

How to use MCP Code Checker?

Install via pip from the GitHub repository, then run the mcp-tools-py command-line tool with the required --project-dir argument. Configure it as an MCP server in clients like Claude Desktop or VSCode using the mcp-config tool or manual JSON configuration. No additional setup beyond pointing to a virtual environment where pylint, pytest, and mypy are installed.

Key features of MCP Code Checker

• Runs pylint, pytest, and mypy on Python projects
• Auto-detects source and test directories from pyproject.toml
• Customizable parameters for each tool (extra args, markers, strict mode)
• Structured JSON logging with function call tracking
• Output formatted as actionable prompts for AI assistants
• Scoped to a single project directory for security

Use cases of MCP Code Checker

• AI‑assisted code review that triggers linting and type checking
• Automated test execution and failure analysis within a chat interface
• Continuous integration style checks triggered from an MCP client
• Developer debugging with detailed logs of each tool execution

FAQ from MCP Code Checker

How does this differ from giving an AI assistant bash access?

The server provides a more controlled environment: only pylint, pytest, and mypy can be run, all operations are confined to the specified project directory, and output is limited and formatted to reduce context consumption. This enhances security and transparency compared to a general-purpose bash MCP tool.

What are the runtime requirements?

The server requires Python and the tools (pylint, pytest, mypy) to be installed in a virtual environment. The --python-executable or --venv-path must point to that environment, not the project’s runtime venv. Installation is done via pip install git+https://github.com/MarcusJellinghaus/mcp-tools-py.git.

Where does the data live?

All operations are performed on the local file system within the specified --project-dir. No data is sent externally. Temporary files are created during test execution and cleaned up unless --keep-temp-files is used.

What are the known limitations?

Only Python projects are supported, and only the three built-in tools (pylint, pytest, mypy) are available. Additional languages would require separate, dedicated MCP servers.

How is the server accessed and authenticated?

The server uses standard MCP transport with no built-in authentication. It is intended for local or trusted network usage. Configuration is done via the client’s MCP settings file (e.g., claude_desktop_config.json).