Skip to content
plugmcp
DirectorySkillsCategoriesBlog
EN/δΈ­ζ–‡
Sign inSign up
Home/Servers/Safe Ssh

Safe Ssh

I
by io.github.Areso

A secured scoped SSH MCP server for executing safe read-only diagnostic DevOps / SysOps commands

Power your agent

Add this server

The publisher's own command. Not a PlugMCP endpoint.

Local command
uvx safe-ssh-mcp==0.1.15
Provider
Claude Code
claude mcp add io.github.areso-safe-ssh-mcp -- uvx safe-ssh-mcp==0.1.15

Run in a terminal.

Claude Desktop
{
  "mcpServers": {
    "io.github.areso-safe-ssh-mcp": {
      "command": "uvx",
      "args": [
        "safe-ssh-mcp==0.1.15"
      ]
    }
  }
}

claude_desktop_config.json

Step-by-step: add this in Claude

Remote URL or a command on your machine? How the two installs differ

Run on PlugMCP

Host a copy with our keys and logs.

A new server you create. It is not this listing's official endpoint. Use the publisher install when you want their URL and their credentials.

Launch

Related MCP

View more
  • FirecrawlπŸ”₯ Official Firecrawl MCP Server - Adds powerful web scraping and search to Cursor, Claude and any other LLM clients.
  • N
    NotionToken-auth Notion MCP: pages, databases, blocks, comments, files. Runs headless in CI and agents.
  • Context7Context7 Platform -- Up-to-date code documentation for LLMs and AI code editors

Related skills

View all
  • reaReverse engineer anything with agents, from app behavior down to native binaries.
  • rea-changelog-updatePrepare or rewrite REA release changelogs and GitHub release notes from pinned Git history, with verified contributor thanks and Release Please synchronization.
  • rea-tool-designDesign or change REA investigation tools, CLI/MCP contracts, provider capabilities, and Evidence semantics.

A secured scoped SSH MCP server for executing safe read-only diagnostic DevOps / SysOps commands

Safe SSH MCP Server

A secure and scoped SSH MCP server for executing read-only diagnostic commands over SSH.
In this project, "safe" refers specifically to host safety: the server is designed to prevent modifications to the remote system and reduce the risk of operational harm. It does NOT attempt to guarantee that command output cannot be misused by external agents.

Badges

Snyk Vulnerability Database report - security check

Overview

The core philosophy behind this MCP server is safety first. Instead of providing an AI agent with an unrestricted bash shell, this server exposes only carefully curated, read-only commands for system diagnostics and monitoring.

Available Tools (Partial List)

  1. get_disk_free : df -h
  2. get_disk_usage : find <path> -mindepth 1 -maxdepth 1 -exec du -sh -- {} + 2>/dev/null | sort -rh | head -n 20
  3. get_dmesg : dmesg
  4. get_uptime : uptime
  5. get_current_datetime : date
  6. get_distroname_version : cat /etc/os-release
  7. get_systemd_list_all : systemctl list-units --all --no-pager
  8. get_systemd_list_failed : systemctl list-units --state=failed --no-pager
  9. get_systemd_list_timers : systemctl list-timers --no-pager
  10. get_crontab_tasks: crontab -l
  11. get_systemd_status : systemctl status {daemon}
  12. get_top : top -b -n 1 -c

Click to open full list of available tools

Example of invocation

okay, could u check disk usage on /root/ path with help of safe-ssh-mcp server on a remote myserver.mydomain.pro using root and /Users/myUser/.ssh/id_rsa to login?

Means of connection

The server supports SSH account-password authentication and private-key authentication. Configure the private-key passphrase source in mcp_config.ini:

# ENV_VAR (SSH_KEY_PASSPHRASE) or KEYCHAIN (ssh-agent)
ssh_key_passphrase_method=ENV_VAR
  1. SSH account password: provide password to the MCP tool. Do not use this for a private-key passphrase.

  2. Unprotected SSH key: use key_path. The default ENV_VAR setting also supports unprotected keys; SSH_KEY_PASSPHRASE is not required.

  3. Passphrase-protected SSH key from an environment variable: leave ssh_key_passphrase_method=ENV_VAR, set SSH_KEY_PASSPHRASE in the environment that starts the MCP server, and provide key_path.

    export SSH_KEY_PASSPHRASE='your-key-passphrase'
    safe-ssh-mcp
    
  4. macOS Keychain / SSH agent: set ssh_key_passphrase_method=KEYCHAIN, then unlock the key in the same user session that starts the MCP server. The server receives SSH signatures from ssh-agent; it does not read the key passphrase or the key_path.

    ssh-add --apple-use-keychain ~/.ssh/id_ed25519
    
    # Optional: make the loaded identity expire after one hour.
    ssh-add -t 1h ~/.ssh/id_ed25519
    
    # Remove one identity, or clear every loaded identity.
    ssh-add -d ~/.ssh/id_ed25519
    ssh-add -D
    
  5. Cursor launched from the macOS Dock: use KEYCHAIN (recommended). You can run ssh-add from any Terminal window; it updates the SSH agent for your macOS login session, so Cursor does not need to be started from that Terminal.

    ssh-add --apple-use-keychain ~/.ssh/id_ed25519
    ssh-add -l
    

    After changing mcp_config.ini, restart the MCP server in Cursor (or fully quit and reopen Cursor). When invoking a tool in KEYCHAIN mode, omit key_path; the agent selects from its loaded identities.

    If you must use ENV_VAR, set it in the macOS launchd session before opening Cursor from the Dock, then fully quit and reopen Cursor:

    launchctl setenv SSH_KEY_PASSPHRASE 'your-key-passphrase'
    

    This value lasts only for the current login session, and putting a passphrase directly in a command may save it in shell history. Prefer KEYCHAIN.

For KEYCHAIN, the MCP server must inherit SSH_AUTH_SOCK from that user session. This is normally automatic for a stdio server started from a desktop application. ENV_VAR stores the passphrase in the server process environment, so restrict access to the process and avoid putting it in source control.

Changelog

Project Contents

  1. mcp_ssh.py - the SSH MCP server
  2. mcp_config.ini - the server's config
  3. check_health_sse.py - check the server's tools and either it's up (sse transport)
  4. check_health_stdio.py - check the server's tools and either it's up (stdio transport)

Compatibility

Tested manually on Python3.11 running on MacOS against remote Ubuntu server
Autotested with Python 3.11 3.12 3.13 3.14

The MCP Registry

mcp-name: io.github.Areso/safe-ssh-mcp

The License

This project is licensed under the GNU AGPLv3 License.

Why AGPL?

This server acts as core infrastructure and contains no business logic. By using the AGPL license, we ensure that any security improvements, bug fixes, or new diagnostic tools added to the server are shared back with the open-source community.

Note for Client Developers

Because MCP clients communicate with this server via standard Inter-Process Communication (IPC) or network protocols (like HTTP/SSE), the AGPL license does not "infect" or restrict the client applications connecting to it.
You can safely connect proprietary, closed-source, or permissively licensed (e.g., MIT, Apache 2.0) AI agents to this server without violating the license terms.

How do I install Safe Ssh?
Copy the stdio config for Cursor, Claude Desktop, Claude Code, or VS Code on this page. It uses the publisher's install, not a PlugMCP proxy.
Does Safe Ssh need an API key?
No required environment variable is listed on this record.
What transport does Safe Ssh use?
Safe Ssh runs locally over stdio.
When was Safe Ssh last checked?
Last checked 2026-10-16. A secured scoped SSH MCP server for executing safe read-only diagnostic DevOps / SysOps commands

No tool list yet. Remote servers get one after the first health check.

plugmcp

Plug any tool into any agent. MCP servers for people who do not write servers. Gateway for volume, n8n for apps.

  • github.com/TODO
  • x.com/TODO
  • linkedin.com/company/TODO

PlugMCP, 2026

Product
  • Browse
  • Install
  • Host your own
  • Voice mode
  • Browser
Directory
  • All MCP servers
  • Agent skills
  • Categories
  • New
  • Voice
  • Submit a server
Developers
  • Docs
  • Blog
  • What is MCP
  • What is WebMCP
  • Status
Company
  • About
  • Contact
  • Terms
  • Privacy
  • DPA