📌Overview

Purpose: The MCP Tools framework aims to extend the capabilities of Large Language Models (LLMs) by providing standardized access to file systems and command execution tools.

Overview: The Model Context Protocol (MCP) facilitates a standard method for LLMs to connect with various data sources and tools, akin to how USB ports function for peripherals. The MCP Tools project implements a FastMCP server that allows LLMs like Claude to safely interact with local systems through a well-defined interface.

Key Features:

execute_shell_command: Enables safe execution of shell commands, returning results for further processing. This empowers LLMs to perform real-time operations on the local system.
show_file: Allows users to view contents of files while specifying line ranges, facilitating easier browsing and debugging of files.
search_in_file: Utilizes regular expressions to search for patterns within files, enhancing text analysis capabilities.
edit_file: Offers precise editing capabilities including string replacements and line operations, enabling users to make targeted modifications to files.
write_file: Supports writing or appending new content to files, useful for logging or data updates.

MCP Tools

A custom Model Context Protocol (MCP) server implementation that provides file system and command execution tools for Claude Desktop and other LLM clients.

What is the Model Context Protocol?

The Model Context Protocol (MCP) is an open protocol that standardizes how applications provide context to Large Language Models (LLMs). Much like a USB-C port provides a standardized way to connect devices to peripherals, MCP provides a standardized way to connect AI models to different data sources and tools.

This project implements a FastMCP server with several useful tools that enable Claude and other LLMs to interact with your local file system and execute commands. It extends LLMs' capabilities with local system access in a controlled way through well-defined tool interfaces.

Key Benefits of MCP

Standardized Integration: MCP provides a growing list of pre-built integrations that your LLM can directly plug into.
Vendor Flexibility: Easily switch between LLM providers and vendors (Claude, GPT-4o, Gemini, etc.).
Security: Best practices for securing your data within your infrastructure.
Tool Exposure: Encapsulate existing tools and make them accessible to any MCP-compatible LLM client.

Features

The MCP server provides the following file system and command execution tools:

execute_shell_command: Execute shell commands and get stdout/stderr results.
show_file: View file contents with optional line range specification.
search_in_file: Search for patterns in files using regular expressions.
edit_file: Make precise changes to files with string replacements and line operations.
write_file: Write or append content to files.

MCP Architecture

MCP follows a client-server architecture:

Hosts: LLM applications (like Claude Desktop or IDEs) that initiate connections.
Clients: Maintain 1:1 connections with servers, inside the host application.
Servers: Provide context, tools, and prompts to clients (this project implements a server).

Prerequisites

Python 3.10 or higher.
An MCP-compatible client (Claude Desktop, or any other client that supports MCP).

Installation

Install uv (uv installation guide)
Clone this repository or download the source code.
Run uv run mcp install to install the MCP server.
Run which uv to get the absolute path to the uv executable.
Update your MCP server configuration in Claude Desktop to use the absolute path to the uv executable.

Example MCP server configuration:

{
  "globalShortcut": "",
  "mcpServers": {
    "zbigniew-mcp": {
      "command": "/Users/zbigniewtomanek/.local/bin/uv",
      "args": [
        "run",
        "--with",
        "mcp[cli]",
        "--with",
        "marker-pdf",
        "mcp",
        "run",
        "/Users/zbigniewtomanek/PycharmProjects/my-mcp-tools/server.py"
      ]
    }
  }
}

Usage

Connecting from Claude Desktop

Open Claude Desktop.
Connect to the MCP server using the identifier zbigniew-mcp.

Note: While this implementation focuses on Claude Desktop, MCP is compatible with any MCP-compatible tool or LLM client, providing flexible integration.

Available Tools

execute_shell_command

Execute shell commands safely using a list of arguments:

execute_shell_command(["ls", "-la"])
execute_shell_command(["grep", "-r", "TODO", "./src"])
execute_shell_command(["python", "analysis.py", "--input", "data.csv"])
execute_shell_command(["uname", "-a"])

show_file

View file contents with optional line range specification:

show_file("/path/to/file.txt")
show_file("/path/to/file.txt", num_lines=10)
show_file("/path/to/file.txt", start_line=5, num_lines=10)

search_in_file

Search for patterns in files using regular expressions:

search_in_file("/path/to/script.py", r"def\s+\w+\s*\(")
search_in_file("/path/to/code.py", r"#\s*TODO", case_sensitive=False)

edit_file

Make precise changes to files:

# Replace text
edit_file("config.json", replacements={"\"debug\": false": "\"debug\": true"})

# Insert at line 5
edit_file("script.py", line_operations=[{"operation": "insert", "line": 5, "content": "# New comment"}])

# Delete lines 10-15
edit_file("file.txt", line_operations=[{"operation": "delete", "start_line": 10, "end_line": 15}])

# Replace line 20
edit_file("file.txt", line_operations=[{"operation": "replace", "line": 20, "content": "Updated content"}])

write_file

Write or append content to files:

# Overwrite file
write_file("/path/to/file.txt", "New content")

# Append to file
write_file("/path/to/log.txt", "Log entry", mode="a")

fetch_page

Fetch the contents of a web page to a PDF (requires chromium installed), then parse it to markdown using local LLMs:

fetch_page("https://example.com")

Transport Mechanisms

MCP supports multiple transport methods for communication between clients and servers:

Standard Input/Output (stdio): Uses standard input/output for communication, ideal for local processes.
Server-Sent Events (SSE): Enables server-to-client streaming with HTTP POST requests for client-to-server communication.

This implementation uses a local MCP server that communicates via text input/output.

Extending with Your Own Tools

You can extend this MCP server by adding new tools with the @mcp.tool decorator. Follow the pattern in server.py to create new tools that expose additional functionality to your LLM clients.

Related Projects

langchain-mcp-adapters: Use MCP with LangChain.
MCP-Bridge: Map MCP tools to OpenAI's format.

Security Considerations

The MCP server provides Claude with access to your local system. Consider the following:

The server executes shell commands as your user.
It can read, write, and modify files on your system.
Limit access to specific directories if security is a concern.

my-mcp-server