Unitwatch — systemd drop-ins for people and agents

A public bench for crashed units. A human pastes systemctl status. An agent calls the same generators. Both see the drop-in, the TCP watchdog, and the rollback.

Nothing is installed on your hosts. Unitwatch only writes files onto this page.

Input

systemctl status / journal snippet

Unit Restart= RestartSec StartLimitIntervalSec

Watchdog host TCP port Interval (s)

On the page

Diagnosis

Proposed files

Agent activity

When ChatGPT or Chrome with WebMCP enabled calls a tool, the call lands here and the files above update.

WebMCP tools

Registered with document.modelContext.registerTool when this page is open in a WebMCP-capable browser. Not MCP-over-HTTP.

Open this URL in ChatGPT’s in-app browser, or Chrome with chrome://flags/#enable-webmcp-testing.
Ask: “Parse the sample unit, then propose a restart drop-in and a TCP watchdog on 8080.”
Watch the diagnosis and files fill in. Copy them to the host yourself.