4.10.0 — The machine’s own load
A snapshot as of 4.10.0 (released 2026-08-19). It will go stale as the app moves on — that is expected, and the living reference is the guide each section links to.
npx mulmoterminal@latest
Nothing to configure. The read-out is on as soon as you update. This page is about what the number means, when it turns colour, and how to switch it off.
The load average, beside the usage gauges

The problem. The grid header already told you what was nearly used up — the 5h and 7d windows your Claude subscription shares across every session. What it never told you was whether this machine could take another agent. That question was answered by opening a terminal and typing uptime, which is exactly the thing nobody does before pressing “new terminal”.
Where it is. The grid header, to the right of the usage gauges. It appears on the grid screen only — the screen where you add cells.
What the number is. The 1-minute load average as a percentage of this machine’s cores:
load 191% ← 38.3 runnable processes on a 20-core machine
Percent-of-cores rather than the raw figure, so it means the same thing on a 4-core laptop as on a 20-core desktop, and so it reads in the same unit as the 5h / 7d figures beside it.
| What you see | What it means |
|---|---|
| grey | under 100% — cores to spare |
| amber, at 100% | every core has work queued; one more cell slows the ones already running |
| red, at 200% | everything on the machine is waiting for a core |
Hover it for the raw 1 / 5 / 15-minute figures, the core count and the multiplier — Load average 38.27 / 58.42 / 60.45 — 20 cores (1.9x). The 1-minute number is what you act on; the 15-minute one tells you whether this is a spike or the last quarter of an hour.
Whose machine. The one running the sessions — the host mulmoterminal is started on, not the device you are looking at the browser from. That is the machine your cells actually compete for.
How fresh. Refreshed every 10 seconds, and never older than 60. A failed request holds the last figure through a blip rather than blanking it, but once the window is spent the figure goes: a load average is a number about now, and a stale one is worse than none.
On Windows it shows nothing at all, and that is deliberate — Windows keeps no load average, and 0% would read as “this machine is idle” at the exact moment it cannot be measured.
Turning it off
Settings → Grid header read-outs (a new pane, under Appearance) has the switch. By hand, in ~/.mulmoterminal/config.json:
{ "showLoadAverage": false }
true / false only; on unless you set false. The reference for every global key is the configuration guide.
Also in this release
- The collections pane has one toolbar. The bar used to be drawn only in a directory that declared a shared app, so it came and went as you moved between cells. It is always there now, with expand and close in it, and a preview opens by default. → Features
- A Claude session can be read from the phone. The remote terminal could only show the current screen — claude runs on the alternate screen, so there is no scrollback to fetch. The host now serves claude’s own transcript, so the phone client can page back through the conversation. → Phone
- Shared apps moved to
sharedapp0.13: live views are declarable, a projection carries the contract version it was written under, the form templates ship the page that reads the answers, and publishing can no longer mint a second app when anaidis deleted. The version pages for 4.8.2–4.8.5 carry the procedure; the skill that writes one ismulmoterminal-shared-app. - Fixes. A cell’s title is read from the transcript instead of being generated by a spawned
claude -p(which had rungitcommands in the working repository); the launcher no longer comes up empty when the backend is still restarting; Codex’s “OR RESUME HERE” lists your own conversations rather than subagent rollouts; and HTML outsideartifacts/opens in the Canvas again underyarn dev.
The full list, with the reasoning behind each, is in the changelog.