StatBar
Sponsor Download
npm version License: MIT macOS 15.0+

System + LLM telemetry in your menubar

Live CPU, GPU, memory, and Apple Silicon telemetry. Plus tokens/sec, KV cache, and context usage for Ollama, llama-server & LM Studio.

brew install coderhd/tap/statbar
npx statbar
curl -L -o ~/Downloads/StatBar.app.zip "https://github.com/coderhd/system-stats-menubar-app/releases/download/v1.1.0/StatBar.app.zip"
git clone https://github.com/coderhd/system-stats-menubar-app && cd system-stats-menubar-app && ./scripts/build-app.sh
What it shows

Everything you need to watch a running LLM

Multi-segment glyph MENUBAR

CPU, GPU, memory and LLM activity as stacked tag + bar segments in one status item — re-rasterized only when a value visibly changes.

LLM server intelligence LLM

Auto-detects llama-server, Ollama and LM Studio. Shows tokens/sec, per-slot context usage, KV cache usage and model RAM live.

Hardware telemetry SYSTEM

Efficiency vs performance core usage, Metal GPU utilization and in-use memory, and a wired / active / compressed / free RAM breakdown.

Native Apple visuals NATIVE

Built with SwiftUI and AppKit. Uses NSVisualEffectView materials, so it matches Light, Dark, Clear Glass and Tinted Glass wallpapers.

Zero idle overhead ZERO

A single lightweight Swift process. When your LLM server is offline the segment auto-hides and StatBar runs as a pure system monitor.

Snapshot mode CLI

Print a single telemetry snapshot straight to stdout with `statbar --once` — scriptable for log tails, CI dashboards or tmux status lines.

Three Ways to Monitor

Every interface at your fingertips

HUD POPOVER

Left-click for the full telemetry HUD

A single click drops down a comprehensive, real-time macOS telemetry dashboard engineered specifically for developers running local LLMs and high-compute workloads.

E/P-Core & Metal GPU Telemetry
Monitor Efficiency vs Performance cores independently with live Metal GPU load, memory footprint, and live sparkline histories.
4-Way Granular Memory Pool
Segmented memory visualization breaking down Active, Wired, Compressed, and Available RAM to gauge system memory pressure instantly.
Live LLM Inference Velocity
Real-time token generation speed (t/s), prompt evaluation rates, total token counts, and active context window utilization.
Left-click Popover HUD Right-click Settings & Quit Native Swift / macOS 15+
CPU
E-cores
23%
P-cores
1%
GPU
Util
3%
Memory 288.7 MB
Memory
Used
5.21 GB / 8 GB
Active 2.17 GB
Wired 1.43 GB
Compressed 1.61 GB
Available 2.79 GB
Load 1.64
Up 7h 36m
Nominal
llama.cpp LFM2.5-2.6B-Q6_K.gguf
idle
127.0.0.1:8080
Context
2,227 / 12,032
Prompt speed 81 t/s
Generate speed 21 t/s
Prompt tokens 171
Generated tokens 2,056
MENUBAR GLYPH

Glanceable at all times

A compact 4-segment animated glyph sitting in your macOS status bar. Tracks CPU, GPU, RAM, and LLM load without taking valuable screen space.

TERMINAL COMPANION

Headless snapshot mode

Need a quick terminal check or script integration? Run statbar --once to output clean CPU, GPU, RAM, and model telemetry directly to stdout.

statbar --once
$ statbar --once
CPU: 34% E: 28% P: 41%
GPU: 12%
MEM: 6.1 GB / 8 GB
LLM: Bonsai-27B @ 18.2 tok/s
Load: 2.91 Uptime: 1d 4h 28m
$
Lightweight & Native

Under 0.1% CPU. Pure native Swift.

No Electron, no web runtime, and zero battery drain when idle. Built specifically for developers running local LLMs on macOS.

Download v1.1.0 Star on GitHub Sponsor
More from CoderHD

Other projects worth a look

DIYPC

AI-powered PC building assistant

Lekhan

Collaborative local-first editor

VahanSaathi

Vehicle management SaaS for dealerships