SentinelX

SentinelX is a Pensa Software product that lets an AI assistant operate servers through an allowlisted, auditable shell. A lightweight agent runs on each Linux, macOS, or Windows host and connects to a hub over a single outbound WebSocket, so no inbound ports are required. The agent is Apache 2.0-licensed.

What It Is

SentinelX is an MCP (Model Context Protocol) server and host agent for remote server operations. After installing the agent and enrolling the host, you connect an MCP client such as ChatGPT, Claude.ai, VS Code, Cursor, or Cline to the hub's connector URL once per account. The assistant can then work with all your enrolled hosts in plain language, for example restarting nginx, checking disk space, tailing an error log, or listing which services are down across the fleet.

Setup Path

On Linux and macOS, a single curl command builds a virtualenv, registers a service (systemd or launchd), writes a starter config, and starts enrollment. Windows uses a PowerShell installer, either as a service in an elevated shell or as a per-user install without admin rights. The installer prints an enrollment URL; you sign in with Google at the hub, enroll a new host, and paste the token back into the terminal. The host then appears in the dashboard.

Tools and Allowlist Model

Per host, the AI gets a fixed set of tools: host state, capabilities, allowlisted command execution, structured file edits, temporary bash or Python scripts, service start/stop/restart/reload, file upload, and a multi-host fleet view. Everything is deny-by-default. A YAML config on the host defines which commands (prefix-matched), services, and editable paths are permitted. The site states that the allowlist, not the model's behavior, is the trust boundary, and that blocked attempts are refused and logged.