SSH-MCP

MCP server that enables LLMs to execute commands on remote servers via SSH.

Overview

This project provides a Model Context Protocol (MCP) server that allows Large Language Models to:

• Connect to remote SSH servers
• Execute commands on those servers
• Retrieve the results

Configuration

Create a configuration file at ~/.ssh-mcp-config.yaml with your SSH connection details:

connections:
  server1:
    hostname: example.com
    port: 22
    username: user
    auth_method: key  # or password
    key_path: ~/.ssh/id_rsa  # only needed for key auth
    
  server2:
    hostname: another-server.com
    port: 2222
    username: admin
    auth_method: password
    password: ${SSH_PASSWORD}  # Use environment variable

defaults:
  timeout: 30  # seconds
  max_output_size: 1048576  # 1MB
  allowed_commands:
    - ls
    - cat
    - grep
    # Add more allowed commands as needed

Installation

# Clone the repository
git clone https://github.com/yourusername/ssh-mcp.git
cd ssh-mcp

# Install dependencies with uv
uv add "mcp[cli]" paramiko pyyaml

# Install the package
uv add --editable .

Usage

Run the MCP server:

mcp dev ssh_mcp/server.py

Docker Usage

For containerized deployment, see the comprehensive Docker Guide.

Quick start:

# Build and run
docker build -t ssh-mcp .
docker run -it --rm \
  -v ~/.ssh:/home/app/.ssh:ro \
  -v ~/.ssh-mcp-config.yaml:/home/app/.ssh-mcp-config.yaml:ro \
  ssh-mcp

Security Considerations

• SSH-MCP implements command allowlisting to prevent arbitrary command execution
• Sensitive information like passwords should be provided via environment variables
• The configuration file should have appropriate permissions (chmod 600)