🔗 RemoteShell MCP

A Model Context Protocol (MCP) server that enables LLMs to securely manage and execute commands on remote SSH servers.

✨ Features

• 🔐 Secure credential management - Save SSH server profiles once, no need to retype credentials
• 💻 Remote command execution - Execute non-interactive shell commands remotely
• 📁 File operations - Upload/download files via SFTP
• 🤖 LLM-powered - Built with FastMCP and Paramiko

🚀 Installation

RemoteShell is a stdio MCP server distributed on PyPI. Run it directly with uvx — no install step required. Then register the server with your AI coding agent:

Claude Code

claude mcp add remoteshell --scope user -- uvx remoteshell-mcp

Codex

codex mcp add remoteshell -- uvx remoteshell-mcp

Other agents — click yours to expand 👇

cursor mcp add remoteshell -- uvx remoteshell-mcp

{
  "mcpServers": {
    "remoteshell": {
      "command": "uvx",
      "args": ["remoteshell-mcp"]
    }
  }
}

opencode mcp add

mcp_servers:
  remoteshell:
    command: "uvx"
    args: ["remoteshell-mcp"]

{
  "mcp": {
    "servers": {
      "remoteshell": {
        "command": "uvx",
        "args": ["remoteshell-mcp"]
      }
    }
  }
}

{
  "mcpServers": {
    "remoteshell": {
      "command": "uvx",
      "args": ["remoteshell-mcp"]
    }
  }
}

📖 Usage

Getting started is easy! You can either:

1. 🤖 Let the LLM configure for you - Simply tell the LLM your host, username, password, etc., and ask it to set up the server configuration
2. ⚙️ Manual configuration - Directly edit the configuration file at ~/.config/remoteshell/hosts.json

💾 Configuration & Storage

RemoteShell securely stores your server configurations in:

~/.config/remoteshell/hosts.json

🔧 Configuration Management

• LLM-managed: The LLM automatically manages this file using save_server and remove_server tools
• Manual editing: You can also directly edit the JSON file for advanced configurations

📋 Example Configuration

{
  "version": 1,
  "servers": {
    "production-server": {
      "host": "1.2.3.4",
      "user": "root",
      "port": 22,
      "auth_type": "password",
      "password": "your_secure_password",
      "last_connected": null
    },
    "staging-server": {
      "host": "staging.example.com",
      "user": "ubuntu",
      "port": 22,
      "auth_type": "private_key",
      "private_key": "~/.ssh/id_rsa",
      "last_connected": "2025-01-01T00:00:00+00:00"
    }
  }
}

🔒 Security Note

On POSIX systems, protect your configuration file:

chmod 600 ~/.config/remoteshell/hosts.json

🛠️ Available Tools

RemoteShell provides the following MCP tools for remote server management:

📋 list_servers()

Purpose: Display all saved server profiles with their connection status and last activity.

When to use:

• User asks to "connect to server" or "show machines"
• No specific connection_id is provided
• Need to see available servers

Example: "Show me which servers I have configured" → Returns list of all saved servers with online status

💾 save_server(connection_id, host, user, auth_type, credential, port)

Purpose: Create or update a server profile with authentication credentials.

Parameters:

• connection_id: Unique identifier for the server (e.g., "production", "staging")
• host: Server hostname or IP address
• user: SSH username
• auth_type: "password" or "private_key"
• credential:
  • For password: Plain text password string
  • For private_key: File path (e.g., ~/.ssh/id_rsa) or PEM key content
• port: SSH port (optional; defaults to 22 and keeps the existing saved port if omitted)

When to use:

• Adding a new server configuration
• Updating credentials after authentication failure
• Changing server connection details

🗑️ remove_server(connection_id)

Purpose: Permanently delete a server profile from storage.

When to use:

• User explicitly requests to remove or forget a server
• Server is no longer accessible or needed

⚠️ Warning: This action cannot be undone

⚡ execute_command(connection_id, command)

Purpose: Execute non-interactive shell commands remotely and return results.

Returns: stdout, stderr, and exit_code

When to use:

• Running system commands, scripts, or utilities
• Checking server status, disk usage, process lists
• File operations, package management, etc.

When NOT to use:

• Interactive programs (vim, htop, top)
• Commands requiring manual input ([Y/n] prompts) - unless using flags like -y

Example: execute_command(connection_id="production", command="df -h")

📤 upload_file(connection_id, local_path, remote_path)

Purpose: Upload files from your local machine to a remote server via SFTP.

Parameters:

• local_path: Path to the file on your local machine
• remote_path: Destination path on the remote server

Notes:

• If remote_path is a directory, the original filename is preserved
• If local_path is omitted, server selects a default and returns it in response

📥 download_file(connection_id, remote_path, local_path)

Purpose: Download files from a remote server to your local machine via SFTP.

Parameters:

• remote_path: Path to the file on the remote server
• local_path: Destination path on your local machine

Notes:

• If local_path is omitted, defaults to: ~/.config/remoteshell/downloads/<connection_id>/<basename>

🧪 Development

Local Development Setup

For local development, use this MCP configuration:

{
  "mcpServers": {
    "remoteshell": {
      "command": "uv",
      "args": ["--directory", "/absolute/path/to/remoteShell-mcp", "run", "remoteshell-mcp"]
    }
  }
}

Running Tests

uv run pytest