---
title: "Hardware"
id: "2223"
type: "page"
slug: "hardware"
published_at: "2026-10-19T23:57:52+00:00"
modified_at: "2026-10-22T16:38:23+00:00"
url: "https://xedant.com/agents/multiagent/docs/hardware/"
markdown_url: "https://xedant.com/agents/multiagent/docs/hardware.md"
excerpt: "Hardware is the live load screen for the machines behind your factories. It lives in…"
---

# Hardware

[https://xedant.com/agents/multiagent/docs/hardware/.md](https://xedant.com/agents/multiagent/docs/hardware/.md)

Hardware is the live load screen for the machines behind your factories. It lives in **Management → Hardware** and answers one question at a glance: which machine is warm, which is busy, and where the disk went.

## What the Screen Shows

The screen reports CPU, memory, disks, load average and uptime for every server. The numbers follow the Docker-panel convention: **CPU and memory are the sums over that machine’s containers** — the same figures a Docker stats panel shows — while disks, load average and uptime come from the machine itself. That is the load your deployments actually place, not the noise of the host’s other tenants.

## The Machine List

The list shows one row per machine: the **address**, linked to the machine’s detail page; **chips for its factories**, each chip opening that factory’s page; and three columns — the containers’ **CPU share**, their **memory share**, and the **disk usage**. The memory and disk cells show the percentage with the “used of total” breakdown beside it in a smaller type — for example, 42% and “30 GB of 64 GB”.

Each column fills independently as that machine’s answer returns — a cell that has no answer yet shows a muted grey dash, never a spinner. The list **refreshes every 15 seconds** while the page is open, one probe per visible machine per round, with the values replaced in place; a failed round keeps the last known numbers, and a hidden browser tab skips the rounds entirely.

## The Machine Detail Page

Clicking a machine’s address opens its detail page: **cores one by one**, **memory and swap**, **each disk** with its used and total, **load average**, **uptime**, the list of the factories on this machine, and three trend charts. The charts gain one point per successful 15-second round — but the points accumulate **in the browser only**: nothing is stored on the machine or in MultiAgent, and leaving the page drops them. The time axis is honest about it — a hidden tab that skipped its rounds shows the gap.

## Fleet Scope

The fleet selected in the header narrows the screen: only machines hosting that fleet’s factories are listed — factories whose agents are fleet members and factories assigned to the fleet alike — and every row’s chips, as well as the machine page’s factory list, show only that fleet’s factories. The “All” scope renders the whole estate.

## Jumping in from a Factory

The **“Hardware statistics”** icon in a factory card header — right next to the factory’s own load figures — opens the page of the machine that factory lives on.

[← Back to documentation](/agents/multiagent/docs)