Show HN: Gumpbox – Let your AI agents work on your remote deployments securely (www.gumpbox.com)

Gumpbox is a native macOS app that embeds an MCP (Model Context Protocol) server to let LLM agents (e.g., Claude, ChatGPT) interact with and manage remote Linux infrastructure through conversation. It provides a full ops UI—multi-tab SSH (up to 10 concurrent sessions per server), dual-pane SFTP, SSH tunneling (local/remote/SOCKS), real-time CPU/memory/storage/network charts, systemd/service management, user/group browsing, visual cron editor, and command templates with variable substitution. All SSH commands are logged with execution time, exit codes and outputs for auditing; you can save and re-run labeled commands, monitor tunnels and connections from a menubar app, and organize multiple servers and environments. The MCP server runs locally on your Mac and brokers AI agent actions to your fleet, while traditional admin tools remain available. For the AI/ML community this lowers the friction for MLOps and AI-driven automation: teams can instruct agents to deploy models, restart services, transfer files, or debug systems conversationally and reproduce workflows from the command history. The audit trail, command templates, and system metrics make it easier to automate and validate deployment workflows, but also underscore the need for careful access controls when granting LLMs operational privileges. Gumpbox positions itself as a secure, offline broker for agent-to-infrastructure workflows with a one-time purchase model (no subscription).