macOS Task Manager with Hardware Monitoring

macOS Task Manager with Hardware Monitoring | Generated by AI

Home 2026.07

Question: What’s JOHN-decm/MacOS-TSKMGR GitHub project?

Answer:

MacOS-TSKMGR GitHub repository is a native macOS Task Manager clone inspired by Windows Task Manager.

Think of it as:

Windows Task Manager UX + macOS native APIs + hardware monitoring.

The author is building a replacement-style utility for macOS, not a direct clone of Apple’s Activity Monitor.

What it does

It provides pages similar to Windows Task Manager:

Processes
Performance graphs
App history
Startup programs
Users
Details
Services

It can show:

CPU usage
Memory
Disk
Network
GPU
NPU
Temperature / cooling information

and actions like:

End task
Restart Finder-like processes
Run new task
Always-on-top window
Refresh-rate switching

Interesting engineering parts

The interesting part is not the UI; it is the hardware telemetry.

1. macOS system monitoring

It uses native macOS interfaces:

sysctl
host_statistics
proc_pidinfo
proc_pid_rusage
IOKit
CoreWLAN
CGWindowList
launchctl

instead of Electron/Tauri/web UI frameworks.

Architecture:

SwiftUI
   |
   +-- AppKit
   |
   +-- System collectors
          |
          +-- CPU
          +-- Memory
          +-- GPU
          +-- Network
          +-- Sensors
          +-- Processes

This is closer to how Apple’s own utilities are built.

2. Apple Silicon NPU / Neural Engine monitoring

This is the most technically interesting piece.

Apple does not expose a nice public API like:

gpu.utilization()
npu.utilization()

Instead the project uses lower-level interfaces:

IOReport
Energy Model
ANS2 Power
AMC Stats Perf Counters

to estimate:

Neural Engine activity
power usage
data movement
memory usage

This is similar to reverse engineering GPU telemetry tools.

3. Cooling / temperature monitoring

It reads:

AppleSMC
SMC sensors
HID temperature nodes
board sensors

to collect:

CPU temperature
GPU temperature
SSD temperature
WiFi temperature
SoC temperature
Logic board temperature
Fan RPM

Tech stack

Native macOS:

Language:
    Swift

UI:
    SwiftUI
    AppKit

Build:
    Xcode

Targets:
    arm64 (Apple Silicon)
    x86_64 (Intel Mac)

Why this project is interesting

For someone doing AI + hardware work, the interesting lesson is:

A lot of “AI workstation monitoring” today is missing this layer.

Imagine a local AI agent:

Agent
 |
 +-- observes:
 |      GPU VRAM
 |      temperature
 |      power
 |      CPU load
 |      memory pressure
 |
 +-- decides:
        kill process
        restart inference server
        change batch size
        reduce context length

This kind of project is basically the foundation for a local AI operating system agent.

For example:

Hermes agent
      |
      |
      +-- vLLM server
      |
      +-- MacOS-TSKMGR style telemetry
      |
      +-- auto optimize inference

Limitations

The difficult parts:

Private Apple APIs

IOReport
AppleSMC

may break after macOS updates.

No official signing yet

The author says it is not currently distributed through Mac App Store because there is no Apple Developer Program signing setup.

Hardware-specific

Apple Silicon generations differ:

M1
M2
M3
M4

have different sensors and Neural Engine behavior.

My take: this is a very good example of a small but deep systems project. It is not huge in LOC, but touches:

OS internals
hardware telemetry
reverse engineering
native UI
performance engineering

Similar spirit to:

nvtop for NVIDIA/Linux
btop
iStat Menus
Windows Task Manager

but focused on modern Mac hardware.

For an AI workstation era, this category (hardware observability + autonomous agents) is likely to become more important.

Back Donate