mcp -fs

This MCP server exposes a file-system oriented set of tools via the MCP interface. It allows you to navigate directories and perform common file operations such as listing, reading, writing, creating, moving, copying, and deleting files and directories. In addition to basic I/O, it provides editing and searching capabilities (edit_file, grep), as well as project analysis (summary) that analyzes Python (.py) and Markdown (.md) files. For batch workflows, you can read or summarize multiple files with read_files and summarize. The project navigation tool work_on helps you quickly move into a project directory and fetch notes from CLAUDE.md to get acquainted with the codebase. For quality assurance, there are linting and formatting utilities (ruff_check, ruff_format). If you need to execute shell commands within the host environment, you can use shell_command, which supports capturing stdout and stderr, with an explicit security warning to prevent arbitrary command execution on untrusted inputs.

Prerequisites:

• Python 3.8+ installed on the host
• Access to install Python packages (pip)

Installation steps:

1. Create a virtual environment (optional but recommended):

• python -m venv venv
• source venv/bin/activate # On Windows use: venv\Scripts\activate

2. Install the MCP server package (if published as mcp-fs-server, adjust to actual package name):

• pip install mcp_fs_server

3. Run the MCP server (as configured in mcp_config):

• python -m mcp_fs_server

4. (Optional) Provide a custom config file via environment variable or CLI if supported by the server, e.g. MCP_FS_CONFIG=path/to/config.json

Notes:

• If your deployment uses a container or orchestration, adapt the run command accordingly (see mcp_config for mapping).
• Ensure the host has permission to read/write the directories you plan to operate on.

Tips and caveats:

• The shell_command tool executes arbitrary shell commands; validate inputs carefully to avoid security risks.
• Use read_files or read_file to retrieve content before applying edits with edit_file to avoid data loss.
• When performing bulk operations (read_files, summarize), be mindful of large files to prevent memory issues.
• The work_on tool is useful for onboarding and quickly reviewing a project's structure and CLAUDE.md notes.
• If you encounter permission errors, verify filesystem permissions for the target paths and consider running the server with elevated privileges if appropriate.