WATCH IT in Chat: Read-Only Monitoring Next to Your Hosts

September 29, 2026 · InfraScout Team

Monitoring tells you that a check is red. It rarely tells you why, and finding out usually means leaving the monitoring console to log on to the server. A first-party integration from ITDESIGN brings WATCH IT into InfraScout, so the AI can read live host and service status and then use InfraScout's agents to look at the host behind the alert, all in the same conversation. It is read-only: the AI can see every check and change none of them. The integration is currently in a test phase.

How It Connects

The integration is an MCP server, so it plugs into InfraScout the same way any other external system does. It is governed by the same tool groups and role tiers as every other tool.

WATCH IT gives the AI the outside view: which hosts are down, which services are critical, and for how long. An InfraScout agent gives it the inside view of the same server: which service stopped, what the event log says, and what filled the disk.

What You Can Ask

The integration offers 12 tools (as of September 29, 2026). They cover live host and service status, events and alerts, host and service groups, and the configuration inventory. A hygiene review tool goes further and looks for monitoring configuration that has drifted or been forgotten, such as hosts that have been down for months with notifications switched off.

A good first question is an overview: how many hosts and services are in each state, and how many problems nobody has handled yet. From there you can narrow it down to a host group, a service type or an application.

What It Looks Like in Practice

The scenario below comes from running the integration against a real WATCH IT environment. Names and identifying details are left out.

The servers behind one application

Asking about the servers that run one business application pulled two long-standing problems into a single answer. On one server, WATCH IT showed Windows services that had been stopped for days. On another, the backup volume was nearly full. Both checks had been critical for several days. Neither was new, and both were easy to miss in a long list of open problems. Filtered down to one application, they became the obvious next thing to look at.

That is where InfraScout's agents take over. The AI can check which services are stopped and why, or what is using the space on the backup volume, and then propose a fix for you to approve.

Keeping monitoring itself tidy

A monitoring setup collects clutter over the years, like hosts that were retired but never removed, and checks that were silenced once and never switched back on. The hygiene review lists them, so you can decide what to clean up. The review only reports. Removing a host or switching notifications back on stays in WATCH IT, done by a person.

Guardrails on the Monitoring Side

The integration uses a read-only role, and it has no tools that change anything. Acknowledging a problem, scheduling maintenance, forcing a recheck and editing the inventory are not available. That protection is built into the server, not a promise in the AI's instructions. A call budget of 200 calls per hour and 1,000 per day caps how much the AI can query, which keeps the load a long conversation adds to your production monitoring small and predictable.

Everything that comes back is treated as data, never as instructions. Free-text fields, such as check output, comments and acknowledgement notes, are often written in German. They are quoted verbatim rather than paraphrased, so you see exactly what the operator or the check wrote.

Ask about one application, not the whole estate

A full list of open problems is a lot to read. Ask about the servers behind one application or host group, then ask the AI to look at the worst one through its agent.

Next Steps

If you run WATCH IT, contact us to connect it to your tenant during the test phase. MCP as the extension layer explains the model behind external integrations, MCP Servers covers registration and the tool allowlist, and Tool Groups is where you decide who gets to use it. To query your identity platform the same way, see Read-Only AI on One Identity Manager.