A high-performance, featherlight, Windows Task Manager inspired software and hardware monitor for macOS. Live multi-engine GPU metrics, non-scrollable command-center dashboard, Mach CPU loads, memory pressure graphs, dedicated hardware deep-dive tabs, per-process telemetry with instant "End Task", live menu bar status, and startup app control.
Download DMG โข Overview โข Screenshots โข Key Features โข GPU Monitoring โข Installation & Build โข Keyboard Shortcuts โข Contributing
Task Manager for macOS brings the familiarity, precision, and power of the Windows Task Manager to the Mac ecosystem. It provides real-time, empirical hardware telemetry and granular software inspection without electron bloat, root daemons, or cloud analytics.
- Zero Web Runtimes: 100% native Swift, SwiftUI, AppKit, Metal, and Darwin Mach kernel APIs. No Electron, Node.js, Chromium, or web servers.
-
Featherlight Overhead: Idles with
< 0.4%CPU usage and~25MBmemory footprint. - Non-Scrollable Single-Window Dashboard: All 5 core performance graphs (CPU, GPU, Memory, Disk, Network) fit into one responsive viewport with zero vertical scrolling.
- Dedicated Hardware Diagnostic Tabs: Clicking any component in the sidebar opens an in-depth diagnostic tab exclusively detailing that component's microarchitecture, pipelines, and top processes.
-
Live Sidebar Sparklines: Real-time mini-graphs rendered directly in the sidebar navigation in strict order: CPU
$\rightarrow$ GPU$\rightarrow$ Memory$\rightarrow$ Disk$\rightarrow$ Network. - Real-Time 1-Second Mac Menu Bar Ticker: The native macOS menu bar status extra streams live telemetry (CPU, GPU, Memory, Disk, or Network) every second in real time.
- True GPU Visualization: Real-time GPU load percentage, renderer utilization, tiler utilization, and allocated/in-use VRAM directly from Apple Silicon & discrete GPUs.
- Memory-First Process Sorting: Processes default-sorted by physical resident memory footprint for instant leak detection, with instant "End Task" process control.
-
100% In-Memory Ephemeral Telemetry: Zero disk I/O writes for metrics; live 60-second rolling ring buffers prevent SSD wear. No SQLite metric rollups
$>60$ seconds. - Single-Tab Non-Scrollable Settings: Cleanly aligned, compact configuration options in one unified, non-scrollable panel fitting neatly within the window viewport.
The non-scrollable Performance Dashboard delivers a unified command-center landing page featuring 60-second real-time charts for CPU, GPU, Memory, Disk, and Network, accompanied by live navigation sidebar sparklines and continuous macOS status bar telemetry.
Dedicated hardware diagnostic views provide in-depth microarchitectural telemetry, core execution distributions, and top resource-consuming processes for each isolated hardware subsystem.
The Processes view provides a real-time inventory of all active macOS applications and background tasks default-sorted by memory footprint with instant "End Task" termination.
The non-scrollable Settings panel offers single-selection macOS menu bar customization, configurable telemetry sampling cadences, visual themes, and complete offline hardware specifications.
- Single-Window Command Center: Displays real-time 60-second interactive charts for all 5 hardware components in a clean 2-row grid fitting completely within the window viewport with zero vertical scrolling:
- Row 1: CPU Card (50% width) and GPU Card (50% width, immediately succeeding CPU).
- Row 2: Memory Card (33% width), Disk Card (33% width), and Network Card (33% width).
- Thermal & Health Ticker: Live thermal throttling status, system uptime counter, and active process count.
- Direct Component Navigation: Clicking any dashboard card or "Details โ" button navigates straight to that component's dedicated in-depth view.
Under Live Hardware Telemetry in the sidebar, selecting any component opens an isolated deep-dive inspection tab:
- CPU: 60s utilization chart, user/kernel/idle time split, core topology, clock speed, cache specs, and top CPU consuming processes.
- GPU (Immediately after CPU): 60s GPU utilization chart, renderer core %, tiler deferred rendering %, in-use VRAM, allocated VRAM, Metal 3 feature set, hardware ray tracing support, and mesh shading capabilities.
- Memory: 60s memory load chart, stacked color-coded Mach VM composition bar (Wired, Active, Compressed, Free), memory pressure status gauge, Mach page size, and top memory-consuming processes.
- Disk: 60s volume space history chart, all mounted APFS/external volume cards with capacity gauges, mount paths, and filesystem attributes.
- Network: 60s dual-line throughput chart (Download โ Cyan, Upload โ Indigo), primary interface status, and detailed interface adapter table with live transfer rates and session traffic.
- Hardware-accelerated mini sparkline graphs stream live in the leftmost navigation sidebar in strict order:
- CPU โ Live utilization % + Real-time sparkline graph
- GPU โ Live GPU load % + Real-time sparkline graph (immediately after CPU!)
- Memory โ Live RAM usage % + Real-time sparkline graph
- Disk โ Live primary disk usage % + Real-time sparkline graph
- Network โ Live throughput rate + Real-time sparkline graph
- Labels ("CPU", "GPU", "Memory", "Disk", "Network") have dedicated column widths ensuring zero text truncation.
- Uses right-aligned fixed-time coordinate mapping with center reference lines and active indicator dots.
- Live Status Bar Label: The macOS top menu bar item displays user-selected real-time hardware telemetry updated continuously every 1 second (e.g.
[CPU icon] 18%or[GPU icon] 41%). - Popup Hardware Monitor: Clicking the menu bar item opens a compact floating HUD displaying live CPU, GPU, Memory, Disk, Network, and Battery stats without opening the main window.
- Common Mode RunLoop: Scheduling on
RunLoop.mainin.commonmode guarantees uninterrupted 1-second updates during user mouse tracking and scrolling.
- Real-time inventory of all running applications and processes with inspectable memory, PID, and architecture (
Apple Silicon/Intel 64-bit). - Memory-First Default Sort: Automatically sorted by resident memory footprint descending to immediately surface memory-heavy applications.
- Live CPU % utilization per process calculated across Mach thread runtime differentials.
- Instant "End Task" action button and context menu with graceful (
NSRunningApplication.terminate) and forced (SIGKILL) termination options. - Instant search filter by application name, bundle identifier, or process PID.
- Live 60-Second Utilization Graph: Continuous GPU engine load time-series rendered via Swift Charts.
- Multi-Engine Breakdown: Overall Device Utilization %, Renderer Core Utilization %, and Tiler Pipeline Utilization %.
- VRAM / Unified Memory Tracking: Real-time allocated system memory and active in-use VRAM.
- Metal Hardware Profile: Dynamic device identification via Apple Metal (
MTLCopyAllDevices).
- 100% Non-Scrollable Layout: Entire settings panel fits comfortably inside the macOS window without vertical scrollbars.
- Single-Selection Menu Bar Widget: Select which hardware component (CPU, GPU, Memory, Disk, or Network) is visible on the macOS top status bar using single-selection radio chips, as only one icon and live telemetry metric can be displayed cleanly at a time.
- Live Menu Bar Preview Ribbon: Immediate simulated preview of the active menu bar status item with a one-click "Reset to CPU" option.
- Telemetry & Sample Rate Control: Custom segmented cadence pills (
0.5s Fast,1.0s Normal,3.0s Eco,Pause) with active interval badge and explicit volatile in-memory ring buffer architecture (zero disk writes). - Appearance & Theme: System, Light, and Dark mode selectors with instant interface adaptation.
- Aligned 2-Column Hardware Specs: Technical specifications (Device Model, Architecture, Processor, Installed RAM, Graphics GPU, OS Build) organized in a clean 2-column grid.
- Desktop Widget Blueprint: See
Documentation/DesktopWidgetGuide.mdfor full architectural patterns and Swift source code to add native macOS Sonoma/Sequoia Desktop Widgets.
You can download the ready-to-run Universal macOS installer directly from the repository:
๐ Download TaskManager-1.0.0-Universal.dmg (v1.0.0)
| Property | Details |
|---|---|
| Package | releases/TaskManager-1.0.0-Universal.dmg |
| Version | 1.0.0 |
| Target Architecture | Universal Binary (arm64 Apple Silicon M1/M2/M3/M4 + x86_64 Intel 64-bit) |
| Minimum OS | macOS 13.0 (Ventura), macOS 14.0 (Sonoma), macOS 15.0 (Sequoia) |
| Package Size | ~5.4 MB (Compressed Read-Only UDZO) |
| SHA-256 Checksum | 657a0f98d5f988c4fc01e029c91be999d22af781d9d946963dabd5a18e810a2e |
- Download TaskManager-1.0.0-Universal.dmg.
- Double-click the
.dmgfile to mount it. - Drag Task Manager into your Applications folder shortcut.
- Eject the disk image and launch Task Manager from Launchpad, Spotlight (โ + Space), or Applications.
Tip
macOS Gatekeeper Notice: Since this open-source build is ad-hoc signed for independent distribution without an Apple Developer ID certificate, macOS may prompt an alert on the first launch. Simply right-click Task Manager.app in /Applications and choose Open, or run the following command in Terminal:
xattr -cr "/Applications/Task Manager.app"To compile both arm64 and x86_64 architecture slices, generate the universal fat binary, assemble the .app bundle, and package a compressed .dmg:
# Clone the repository
git clone https://github.com/K-692/Task-Manager4MacOS.git
cd Task-Manager4MacOS
# Build Universal app and package DMG
./Scripts/package-dmg.shThe output disk image will be created at dist/TaskManager-1.0.0-Universal.dmg and updated in releases/.
To assemble a native .app bundle directly on your Mac:
# For Universal (Apple Silicon + Intel) bundle:
./Scripts/build-universal.sh
# For native host architecture bundle:
./Scripts/build-app.sh
# Open Task Manager
open "dist/Task Manager.app"swift run TaskManager| Shortcut | Action |
|---|---|
| โ + 1 | Navigate to Performance Dashboard (Landing Page) |
| โ + 2 | Navigate to Processes (Default: Memory Footprint) |
| โ + 3 | Navigate to Startup Apps |
| โ + 4 | Navigate to Settings |
| โ + 5 | Navigate to CPU Telemetry Detail View |
| โ + 6 | Navigate to GPU Telemetry Detail View |
| โ + 7 | Navigate to Memory Telemetry Detail View |
| โ + 8 | Navigate to Disk Telemetry Detail View |
| โ + 9 | Navigate to Network Telemetry Detail View |
| โ + K | Open Command Palette (Instant Search & Actions) |
| โ + R | Force Immediate Hardware Telemetry Refresh |
- 100% Offline: Zero network requests, zero telemetry beaconing, zero analytics SDKs.
- Purely In-Memory Telemetry: Telemetry history is strictly volatile (in-memory 60-sample ring buffers).
- User Space Execution: Zero root permissions or kernel extensions required.
Contributions to Task Manager for macOS are welcome! Whether filing bug reports, suggesting new hardware sensors, or submitting pull requests:
- Fork the repository on GitHub.
- Create a branch for your feature (
git checkout -b feature/hardware-sensor). - Commit your changes (
git commit -m 'Add support for Apple Neural Engine telemetry'). - Push to branch (
git push origin feature/hardware-sensor). - Open a Pull Request with architectural explanations.
This project is licensed under the MIT License โ see the LICENSE file for details.



