MacJet

MacJet is a high-performance, developer-first terminal dashboard designed to answer the one question every Mac user asks eventually: "Why does my laptop sound like a jet engine?"

Version 2.0.1 is a complete ground-up rewrite in Rust using ratatui and tokio. The table below uses benchmark_compare (sysinfo, % of one core): a 2026-03-27 run with MacJet --no-ml, 37 samples, 10 s spacing (~6 min wall). In that session, MacJet v2.0.1 averaged ~2.8% CPU vs Activity Monitor ~1.2%; historical Python/Textual v1 figures (14.7% avg CPU, 65 samples, 2 windows) come from docs/benchmarks.md (published alongside the v2.0.1 Rust release). That is roughly ~5× lower average CPU for Rust vs the documented Python baseline. Full methodology and older sessions: docs/benchmarks.md. JSON artifact: benchmark_compare_1774606304.json at repo root (copy to benchmarks/results/ after mkdir -p benchmarks/results if you use the default benchmark output path). Regenerate tables from committed JSON with python3 scripts/benchmark_readme_snippet.py.

⚡ Quick Start

git clone https://github.com/jeppsontaylor/MacJet.git
cd MacJet
cargo install --path .

Then, launch it:

sudo macjet

From GitHub Releases

Each tagged release attaches macjet-macos-universal.tar.gz (fat binary: Apple Silicon + Intel). Extract macjet, place it on your PATH, and clear Gatekeeper quarantine if macOS blocks it: xattr -dr com.apple.quarantine /path/to/macjet.

To sudo or not to sudo?

MacJet is designed to run gracefully with or without root privileges, but sudo unlocks its true power by granting access to Apple's low-level sensors.

With sudo (Recommended):

• 🌡️ Thermal Data: Access CPU/GPU die temperatures and Fan speeds (RPM).
• 🔋 Energy Impact: Accurate, hardware-level energy scoring via powermetrics.
• 🛡️ Full Control: Ability to analyze and manage any process on the system, not just your own user processes.

Without sudo:

• 📊 App Grouping & UI: Full access to the terminal UI, App-centric grouping, and Chrome tab mapping.
• 📉 Basic Metrics: Standard CPU/Memory metrics.
• 🔒 Restricted Control: Can only interact with your own user-level processes. Thermal data and energy metrics will be disabled.

🚀 Performance: Activity Monitor vs MacJet v1 (Python) vs v2.0.1 (Rust)

CPU and RSS are % of one core and MB from benchmark_compare / historical docs. Python v1 is not from the same sampling session as the 2026-03-27 row.

Takeaway: In the benchmark_compare session, both apps show small non-zero idle CPU (measurement noise + background work); MacJet is higher than Activity Monitor on average CPU here but far below the documented Python v1 average (~5× vs 14.7%).

Memory: In that same session, MacJet RSS (~53 MB) was lower than Activity Monitor (~88 MB). Separately, an older documented session (docs/benchmarks.md) reported ~27 MB (v1) — v2 natively trades some static RAM for Tokio, caching, and a richer UI, but uses far less CPU.

✨ Features

🎯 Flight Deck Layout

Five purpose-built views, switchable with 1–5 or Tab:

🧠 Reclaim Engine (Kill List)

A multi-factor scoring engine ranks every process group on a 100-point scale based on: sustained CPU load, memory footprint, memory growth (leaks), process storms, and high wakeups. Target high-score apps to reclaim your system.

🌐 Chrome Tab Mapping

Connects to Chrome's DevTools Protocol to map every renderer PID to its actual website tab title. Stop guessing which "Google Chrome Helper (Renderer)" is drawing 100% CPU. Enable with open -a "Google Chrome" --args --remote-debugging-port=9222.

🤖 Built-in MCP Server

MacJet ships a native Model Context Protocol (MCP) server natively exposing 10 tools to AI Agents (like Claude Desktop).

To use it, just configure your MCP client:

{
  "mcpServers": {
    "macjet": {
      "command": "/Users/YOU/.cargo/bin/macjet",
      "args": ["--mcp"],
      "description": "macOS process monitor — CPU, memory, energy, Chrome tabs, process management"
    }
  }
}

📖 Read the full MCP capabilities in docs/mcp.md

⌨️ Keybindings

🤝 Contributing & Architecture

We welcome contributions! See the architecture doc, benchmarks, and CONTRIBUTING.md to get started with the Rust codebase.

📜 License

MIT © Jepson Taylor