AIOps that shows its evidence

Root-cause analysis over metrics, logs and traces with every conclusion cited and the fix one click away; learned baselines, alert grouping and policies you control.

Book a demoSecurity posture

AI that shows its evidence

An answer you cannot check is a guess. Rapitt’s AI is built so that every conclusion points at the evidence behind it, every suggested fix is something the platform can really do, and the person decides.

It works over what Rapitt already holds - metrics, logs, traces, alarms, operator actions - and, when you ask, over fresh evidence collected on the server itself.

Root cause
A short statement of what failed and why.
Evidence
Each claim cites a log line, a trace or a measurement: [log:n] [trace:id]
Remediation
A button for an action the platform can run, or a manual step when it cannot.
The shape of a root-cause report. Schematic; the wording is not real output.

What the AI layer does

Seven features, one rule: show the work.

Root-cause analysis

Pick a server and a window; Rapitt collects evidence on the server (Windows or Linux, up to five servers in one analysis) and combines it with traces, logs and what it recorded. The result is a branded PDF with cited conclusions and a timeline.

Fixes you can run

Remediations are chosen only from the actions the platform can run on that host: service and host actions, Auto Fix actions, library scripts. Each has a Run button with a confirmation, or a manual step.

Ask Rapitt

Ask about your estate in plain language: which hosts are short of disk, what changed overnight. Answers use only what the asker may see, and it cannot act on a server.

Knowledge store

Upload runbooks and keep past incidents. For a new problem Rapitt retrieves the relevant ones, suggests next steps and cites the source of each. One client’s documents are never used for another client’s server.

Baselines and anomalies

Each host’s normal is learned from its own history. A reading that surprises both the seasonal profile and the forecast, for several readings in a row, raises an anomaly. Smart thresholds follow what each host really does.

Alert grouping

One cause, one message. Hosts behind a failed tunnel are grouped with the tunnel, repeated alarms on one host or one service are combined, and a storm becomes one summary. Nothing is hidden: every alert is still listed.

Recommendations

Right-sizing with the monthly saving, disks that will fill, certificates, stopped services, security fixes. Each shows its evidence and, where there is a button for it, an Apply.

Auto Fix

A policy reads: if this condition holds and nobody acts within N days, run this action. Off by default, with a cooldown, and every run is logged with its outcome.

Not autonomous, on purpose

Rapitt does not claim to run itself. Suggestions need a person; automatic actions are limited to the ones an operator wrote down, and the engine can be switched off with one switch. When a person acts on a condition, the Auto Fix clock restarts.

Because suggestions are checked on the server against what the platform can really run, a convincing but impossible fix never reaches a button.

Related

AIOps questions

Does the AI act on my servers?

Not on its own. Ask Rapitt answers questions and cannot act on a server. A root-cause report proposes fixes only from actions the platform can run, and running one is a button an authorised person presses, through the same audited path as any manual action. Auto Fix is off by default and runs only actions an operator set up.

How do I know the AI is not making things up?

Every conclusion in a root-cause report cites the evidence it was drawn from, and the server keeps only citations that exist in the evidence it sent. A proposed fix that is not in the list of actions the platform can run is dropped and shown as a manual step.

What does the AI see?

Only the evidence collected for that analysis and what Rapitt recorded about the host. Ask Rapitt answers using only what the person asking is allowed to see. The privacy notice names who receives data when AI analysis is used.

Is the anomaly detection deep learning?

No. Baselines are statistical: each host’s own history gives a seasonal profile and a forecast, and a reading must surprise both before it counts. The console says so.

See it on your own systems

A walkthrough runs on your own estate. Reserve your workspace address while you book.

See your own estate in Rapitt

A walkthrough on your own systems takes an afternoon. Reserve your workspace address while you book.

Book a demo [email protected]
  1. Open the Camera
  2. Point it at the code and tap the link that appears.