monmon 0.1.0

macOS silicon monitor (E/P cores, GPU, NPU) with a TUI

monmon

A terminal UI for watching Apple Silicon work: efficiency cores, performance cores, GPU, and the Neural Engine (NPU) — plus the top processes driving load.

One-shot build note: this entire project was generated in a single pass by Claude Opus 4.7 (1M context) running in Claude Code, from the prompt shown at the bottom of this README. No manual edits were made before the first successful run.

Requirements

• macOS on Apple Silicon (M1 or newer)
• Python 3.12+
• uv
• sudo access (Apple's powermetrics requires root)

Install / run

Homebrew (recommended for users)

brew install gavi/monmon/monmon
monmon

From source (for development)

uv sync
uv run monmon

You'll be prompted for your password once — sudo caches it for the session and powermetrics is spawned under sudo -n so it never blocks the TUI.

Optional flags:

uv run monmon --interval 500   # 500 ms samples (default: 1000)

Quit with q or ctrl-c.

What you see

• CPU panel: every E-core and P-core cluster with per-core active-residency bars and current frequency. E-clusters render in cyan, P-clusters in magenta.
• GPU panel: active residency, frequency, and power draw.
• NPU panel: Apple Neural Engine power. powermetrics doesn't expose an "active" counter for ANE, so monmon treats any non-trivial power draw as "in use" and scales the bar against an 8 W ceiling.
• Process table: top processes by CPU from psutil. (powermetrics' task sampler is unreliable across macOS versions, so we use psutil for the "what's running?" view.)

Data source

Everything above comes from a single powermetrics -f plist stream with the cpu_power, gpu_power, and ane_power samplers. Samples are NUL-delimited XML plists; see src/monmon/power.py for the parser.

Origin prompt

This is the exact prompt that produced the project, verbatim:

lets build a mac monitoring system that shows gpu
lets use
uv for package management and tui
we need to see e-cores and p-cores and also gpu and npu to see what is being run
go ahead a build it

Model: Claude Opus 4.7 (1M context) via Claude Code. One-shot — the only follow-up was a cosmetic request to add this section and a CLAUDE.md.