my

This MCP server exposes a local set of tools that let an LLM-based client interact with the host system. It provides file system access and command execution capabilities through a standardized MCP tool interface, enabling tasks such as listing and viewing files, editing content, performing targeted searches, and running shell commands. Clients (like Claude Desktop or other MCP-enabled tools) can connect to this server to request these operations without direct shell access, maintaining a controlled abstraction layer between the model and the host environment.

To use the server, connect your MCP client to the server entry (identified as zbigniew-mcp in Claude Desktop or the corresponding MCP client). Once connected, you can invoke the available tools by name, passing arguments in the structured MCP format. For example, you can run a shell command via execute_shell_command, view a file with show_file, search for patterns with search_in_file, modify content with edit_file, or write data using write_file. Each tool returns structured results (stdout, stderr, line ranges, or applied changes) that your client can present to the user or feed back into the conversation flow.

Prerequisites:

• Python 3.10 or higher
• Access to install and run uv (the Ultralight Python/Vintage runtime used by MCP servers)

Install and run locally:

1. Install uv and ensure it is available in your PATH. See the uv installation guide: https://docs.astral.sh/uv/getting-started/installation/

2. Clone this repository or download the source code to a local directory.

3. Install the MCP server runtime (via uv) using:

   uv run mcp install

4. Verify the uv executable path (absolute path) on your system with:

   which uv

5. Configure Claude Desktop (or your MCP client) to point to the absolute uv path and the server script path (server.py). For example, the server configuration in Claude Desktop should reference the uv invocation and the path to server.py, similar to the included example in the README.

6. Start the server by using your MCP client to connect to the server entry point; the client will load the mcp server over the uv runtime as configured.

Tips and considerations:

• The server runs with your user permissions; commands and file access are subject to your user rights. Be cautious about allowing access to sensitive directories.
• If you modify server.py to add or adjust tools, restart the MCP server so changes take effect.
• If you encounter path issues, prefer absolute paths for both the server script (server.py) and any files accessed by tools.
• Ensure your MCP client supports the MCP tool interface and that the tool names (e.g., execute_shell_command, show_file, edit_file, write_file, search_in_file) match exactly what the server exposes.
• When deploying, consider restricting network exposure and using authentication to protect your local environment.
• For debugging, check the MCP server logs produced by uv and the client-side responses to confirm arguments and results are being passed correctly.