kennss/SiliconScope

Sudoless Apple Silicon system monitor (native SwiftUI GUI) with ANE / Media Engine / memory-bandwidth tracking

SiliconScope – Apple‑Silicon System Monitor

What it is

  • A native macOS 14+ app (SwiftUI) that visualises low‑level hardware metrics on Apple‑Silicon Macs. It goes beyond Activity Monitor by exposing the Neural Engine (ANE), Media Engine, memory‑bandwidth, per‑core (E‑core/P‑core) utilization, GPU stats, power domains, temperatures, battery health and more.
  • Includes a menu‑bar suite so each metric can live as its own icon with a dropdown, and a full‑screen dashboard that can record and replay sessions.
  • New Fleet feature (v4.0) lets you add remote machines – headless Macs, Linux GPU boxes, or cloud VMs – via a tiny agent, showing their metrics in the same dashboard over TLS‑encrypted, token‑authenticated connections.

Key features

  • ANE & Media Engine monitoring – per‑process Neural‑Engine memory, cluster residency, power.
  • Process Inspector – live per‑process CPU split, IPC, power (W), memory, disk I/O, and ANE usage.
  • Record & Replay – dump all metrics to a compact .ssrec file and scrub through the timeline like a DVR.
  • AI workload classifier – tells whether a local LLM run is bandwidth‑bound, compute‑bound, thermally throttled, etc., based on chip‑specific spec ceilings.
  • Menu‑bar metrics – pin any card (CPU, GPU, Memory, Network, SSD, Sensors, Battery) as an independent menu‑bar item with customizable visual style.
  • Fleet dashboard – monitor multiple Macs, NVIDIA/RTX Linux boxes, or headless servers in one view; automatic LAN discovery via mDNS, optional Tailscale support.
  • No sudo required – all data is gathered via private IOReport and other system APIs that work without elevated privileges.
  • Self‑updating – Sparkle‑based auto‑updates; also distributable as a Homebrew Cask.

Typical use cases

  • Developers building on‑device AI (Core ML, Spectalo, etc.) who need to see how the Neural Engine and memory bandwidth are utilized.
  • Power users replacing iStat Menus with a more detailed, Apple‑Silicon‑specific monitor.
  • System administrators who want a unified view of a mixed fleet (Mac minis, Linux GPU servers, cloud containers) without logging into each host.
  • Researchers benchmarking on‑device LLM inference, using the built‑in “tokens / sec” and “tokens / Wh” measurements.

Installation

# via Homebrew (recommended)
brew install --cask siliconscope

Or download the signed DMG from the latest GitHub release, drag the app to Applications, and launch. The app is notarized and runs without Gatekeeper warnings.

Running the Fleet agent

curl -fsSL https://raw.githubusercontent.com/kennss/SiliconScope/main/scripts/install-agent.sh | sh

The script prints a sscope://pair… URL; paste it into Add machine… in the app to pair the remote host.

License: MIT (source code freely reusable under standard MIT terms).

Platform: macOS 14+ on Apple Silicon (M1‑M5). Remote agents also run on Linux (GPU‑focused view) and on headless macOS or Windows (limited metrics).

Why it matters SiliconScope is the first consumer‑grade tool that surfaces per‑process Neural‑Engine usage and real‑time memory‑bandwidth on macOS, filling a gap left by both terminal utilities and commercial products like iStat Menus. Its Fleet capability makes it a practical monitoring hub for heterogeneous compute environments that include Apple‑Silicon devices.

Related

  • Project
    Sentient-OS-Labs/sentient-os

    Sentient OS is an open‑source macOS personal AI that reads your local files, mail, calendar and messages, builds a Markdown knowledge base, and lets you hand off tasks via a shortcut. It can type, click, fill forms (Sidekick), draft replies in your voice (Double‑Tap), and surface proactive suggestions after overnight on‑device processing. You can run the built‑in model locally or plug in any compatible local or hosted LLM, and all raw data stays on your Mac unless you enable optional encrypted cloud sync.

  • Project
    hollance/neural-engine

    A community‑maintained guide that documents how Apple’s on‑device Neural Engine works, which iOS devices have it, how to get Core ML models to run on it, how to detect usage, what layers are unsupported, and tips for work‑arounds—all gathered from the author’s experiments.

  • Project
    IBM/AssetOpsBench

    AssetOpsBench is an open‑source benchmark/framework for building, orchestrating, and evaluating LLM‑based AI agents that operate on industrial asset‑management data (sensors, work orders, failure modes, etc.). It provides domain‑specific MCP tool servers, several ReAct‑style agent back‑ends, 141+ realistic scenarios, and a multi‑dimensional evaluation pipeline used in KDD/AAAI/NeurIPS competitions.

  • Dispatch
    Breadcrumb AI Plugin for macOS: Local Context Management and Recording

    Breadcrumb is a free macOS app that records screen activity and provides AI‑driven context management while keeping all data encrypted and local.

  • Project
    krrishnarraj/clpeak

    clpeak is an open‑source cross‑platform benchmark that measures peak compute throughput, memory bandwidth and latency on CPUs, GPUs and NPUs. It supports many backends (CUDA, ROCm, Metal, oneAPI, Vulkan, OpenCL, ONNX Runtime, Core ML, LiteRT), offers both CLI and GUI, and can export results as JSON for comparison.