Fire in da houseTop Tip:Paying $100+ per month for Perplexity, MidJourney, Runway, ChatGPT and other tools is crazy - get all your AI tools in one site starting at $15 per month with Galaxy AI Fire in da houseCheck it out free

Pandoras-Shell

MCP.Pizza Chef: Zelaron

Pandoras-Shell is an MCP server that allows AI assistants to execute terminal commands directly on your system through a secure shell interface. It provides full system access, enabling execution of any shell command, capturing command output (stdout and stderr), setting working directories, and handling command timeouts. Due to the unrestricted nature of this access, it is strongly recommended to use Pandoras-Shell only in controlled environments such as virtual machines or disposable development systems to mitigate security risks. This server is compatible with any AI assistant supporting the Model Context Protocol (MCP), with tailored setup instructions for Claude Desktop to leverage all MCP features effectively.

Use This MCP server To

Execute shell commands via AI assistants securely Capture and analyze terminal command output Automate system tasks through AI-driven shell access Set working directories for context-specific command execution Manage command execution timeouts to prevent hangs

README

Pandora's Shell

⚠️ IMPORTANT SECURITY WARNING: This MCP server grants AI assistants unrestricted ability to execute terminal commands on your system. Only use in controlled environments like virtual machines (VMs) or development systems you can afford to rebuild.

About

An MCP server that empowers AI assistants to execute terminal commands on your system. Due to the unrestricted access this provides, it's crucial to use this software responsibly and be fully aware of the security risks involved.

Note: This server is compatible with any AI assistant that supports the Model Context Protocol (MCP). The provided configuration and setup instructions are specifically tailored for Claude Desktop, which offers comprehensive support for all MCP features.

Features

  • Execute any shell command with full system access
  • Capture command output (stdout/stderr)
  • Set working directory
  • Handle command timeouts

API

Tools

  • execute_command
    • Execute any shell command and return its output
    • Inputs:
      • command (string): Command to execute
      • directory (string, optional): Working directory
    • Returns:
      • Command exit code
      • Standard output
      • Standard error
    • Features:
      • 5-minute timeout
      • Working directory support
      • Error handling

Installation

Prerequisites

  • Claude Desktop with an active Claude Pro/Enterprise subscription
  • Python 3.10 or higher
  • Git
  • uv (required for package management)

Windows Installation

  1. Install Prerequisites:

    Option A - Using winget (if available on your system):

    winget install python git

    Option B - Manual installation (recommended):

  2. Install uv:

    Open Command Prompt (cmd.exe) as administrator and run:

    powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex"

    If you encounter any issues, you may need to restart your terminal or computer for the changes to take effect.

  3. Clone and set up the project:

    git clone https://github.com/Zelaron/Pandoras-Shell.git
    cd Pandoras-Shell

    Then create a virtual environment. Try these commands in order until one works:

    python -m venv venv

    If that doesn't work, try:

    python3 -m venv venv

    Then activate the environment:

    venv\Scripts\activate
  4. Install dependencies:

    uv pip install mcp
    pip install -e .

Note: If you installed Python from python.org, you'll typically use python. If you installed via winget or from the Microsoft Store, you might need to use python3. Try both commands if one doesn't work.

macOS Installation

  1. Install Prerequisites:

    brew install python git uv
  2. Clone and set up the project:

    git clone https://github.com/Zelaron/Pandoras-Shell.git
    cd Pandoras-Shell
    python3 -m venv venv
    source venv/bin/activate
  3. Install dependencies:

    uv pip install mcp
    pip install -e .

Configuration

Windows

Locate the correct configuration directory - try these paths in order:

  1. %APPDATA%\Claude\ (typically C:\Users\[YourUsername]\AppData\Roaming\Claude\)
  2. %LOCALAPPDATA%\AnthropicClaude\ (typically C:\Users\[YourUsername]\AppData\Local\AnthropicClaude\)

Create or edit claude_desktop_config.json in the correct directory:

{
  "mcpServers": {
    "pandoras-shell": {
      "command": "C:/path/to/cloned/Pandoras-Shell/venv/Scripts/python.exe",
      "args": [
        "C:/path/to/cloned/Pandoras-Shell/src/pandoras_shell/executor.py"
      ],
      "env": {
        "PYTHONPATH": "C:/path/to/cloned/Pandoras-Shell/src"
      }
    }
  }
}
Important Notes for Windows:
  • Use forward slashes (/) in paths, not backslashes (\)
  • Replace [YourUsername] with your actual Windows username
  • File must be named exactly claude_desktop_config.json
  • If both possible config locations exist, try each until successful

macOS

Create or edit ~/Library/Application Support/Claude/claude_desktop_config.json:

{
  "mcpServers": {
    "pandoras-shell": {
      "command": "/path/to/cloned/Pandoras-Shell/venv/bin/python",
      "args": [
        "/path/to/cloned/Pandoras-Shell/src/pandoras_shell/executor.py"
      ],
      "env": {
        "PYTHONPATH": "/path/to/cloned/Pandoras-Shell/src"
      }
    }
  }
}
Important Notes for macOS:
  • Replace [YourUsername] with your actual username
  • You can use $HOME instead of /Users/[YourUsername] if preferred
  • File must be named exactly claude_desktop_config.json
  • The command path should point to the Python interpreter inside your virtual environment (venv/bin/python), not the system Python

After Configuration

  1. Restart Claude Desktop completely (quit/exit, not just close the window).
  2. Click the 🔌 icon to verify the server appears in the "Installed MCP Servers" list.
  3. If the server doesn't appear, check Claude's logs:
    • Windows: %APPDATA%\Claude\Logs\mcp*.log or %LOCALAPPDATA%\AnthropicClaude\Logs\mcp*.log
    • macOS: ~/Library/Logs/Claude/mcp*.log

Security Considerations

This server executes commands with your user privileges. Take these precautions:

  • Use only in VMs or disposable development environments.
  • Never use on production systems or machines with sensitive data.
  • Consider implementing command restrictions if needed.
  • Monitor system access and activity.
  • Keep backups of important data.

Disclaimer: The developers are not responsible for any damages or losses resulting from the use of this software. Use it at your own risk.

Troubleshooting

If you encounter issues:

  1. Check logs:

    • Windows: %APPDATA%\Claude\Logs\mcp*.log or %LOCALAPPDATA%\AnthropicClaude\Logs\mcp*.log
    • macOS: ~/Library/Logs/Claude/mcp*.log
  2. Verify installation:

    • Ensure uv is properly installed and in your PATH.
    • Check that mcp package is installed: pip show mcp.
    • Verify Python version is 3.10 or higher.
  3. Configuration issues:

    • Double-check all paths in claude_desktop_config.json.
    • Verify JSON syntax is valid.
    • Ensure proper path separators for your OS.
    • Confirm config file is in the correct location.
  4. Environment issues:

    • Make sure virtualenv is activated if using one.
    • Verify PYTHONPATH is set correctly.
    • Check file permissions.
  5. Test server manually:

    # First, make sure you're in the Pandoras-Shell directory:
    cd /path/to/cloned/Pandoras-Shell
    
    # For macOS:
    ./venv/bin/python src/pandoras_shell/executor.py
    
    # For Windows:
    .\venv\Scripts\python.exe src\pandoras_shell\executor.py
    
    # The executor will appear to hang with no output - this is normal.
    # It's waiting for connections from Claude Desktop.
    # Use Ctrl+C to stop it.
  6. Connection issues:

    • If you get "Could not connect to MCP server" errors, ensure you're using the virtual environment's Python interpreter in your config file.
    • For macOS: Use /path/to/cloned/Pandoras-Shell/venv/bin/python
    • For Windows: Use C:/path/to/cloned/Pandoras-Shell/venv/Scripts/python.exe

Testing

After setup, try these commands in Claude Desktop:

Can you run 'pwd' and tell me what directory we're in?

or

Can you list the files in my home directory? Which of them are larger than 200 MB?

Pandoras-Shell FAQ

How do I safely use Pandoras-Shell given its unrestricted command execution?
Use Pandoras-Shell only in controlled environments like virtual machines or disposable development systems to avoid security risks.
Can Pandoras-Shell capture both standard output and error streams from commands?
Yes, it captures stdout and stderr from executed shell commands for comprehensive output handling.
Is Pandoras-Shell compatible with all AI assistants supporting MCP?
Yes, it works with any MCP-compatible AI assistant, with specific setup guidance available for Claude Desktop.
How does Pandoras-Shell handle long-running or stuck commands?
It supports command timeouts to automatically terminate commands that exceed specified execution durations.
Can I specify the working directory for commands executed via Pandoras-Shell?
Yes, you can set the working directory to control the context in which commands run.
What precautions should I take before deploying Pandoras-Shell?
Ensure you understand the security implications and restrict usage to safe, isolated environments to prevent unauthorized system access.
Does Pandoras-Shell provide any access control or restrictions on commands?
No, it grants unrestricted shell command execution, so access control must be managed externally.
Can I integrate Pandoras-Shell with multiple LLM providers like OpenAI, Claude, and Gemini?
Yes, since it uses MCP, it can integrate with any LLM provider supporting the protocol, including OpenAI, Claude, and Gemini.