A terminal dashboard that shows you exactly how hard your Mac is working when running Claude agents — so you can tell at a glance whether you have room to spin up more instances or if your machine is already at its limit.

Built with Rust. Single binary (~800 KB), no runtime dependencies.

If you run multiple Claude Code agents in parallel — across different projects, terminals, or background tasks — you've probably asked yourself: can I open one more, or will my Mac start crawling?

cmon answers that question. It gives you a single, always-on view of:

• How much CPU and RAM each Claude process is using
• How much system memory is left before things start swapping
• What other processes (Docker, Colima, Postgres, etc.) are competing for resources
• How much disk space Claude's data is taking up
• Your current network and internet throughput

Open it in a corner terminal, keep it running, and you'll always know when it's safe to launch another agent.

┌──────────────────────────────────────────────────────────────────────┐
│  CMON - Claude Monitor                          Uptime: 2h 34m     │
├──────────────────────────────────────┬───────────────────────────────┤
│  PROCESSES (CPU + RAM)               │  NETWORK I/O                 │
│  PID    Name     CPU%   RAM  Status  │  PID    In       Out         │
│  44326  claude   35.3%  3.3 GB  Run  │  44326  152K/s   3.8MB/s     │
│  80797  claude   17.6%  2.2 GB  Run  │  TOTAL  646K/s   5.0MB/s     │
│  TOTAL           53.3%  5.5 GB       │                              │
├──────────────────────────────────────┤  RAM: 12.4 / 16.0 GB (78%)  │
│  CPU: 23.4% (10 cores)              ├───────────────────────────────┤
│  44326: ▁▁▂▃▅▇████▇▅▃▂▁            │  Name           #   RAM      │
│  Total: ▁▂▃▅▇█████▇▅▃▂             │  claude          3   3.30 GB  │
│                                      │  colima          2   2.14 GB  │
│                                      │  Others        187   1.92 GB  │
├──────────────────────────────────────┼───────────────────────────────┤
│                                      │  DISK / SSD    INTERNET      │
│                                      │  Free: 331 GB  ↓ 1.2 MB/s   │
│                                      │  [========---]  ↑ 340 KB/s   │
├──────────────────────────────────────┴───────────────────────────────┤
│  Last refresh: 19:28:03  |  Tick #42  |  'q' quit  'r' refresh     │
└──────────────────────────────────────────────────────────────────────┘

• Claude process tracking — CPU%, RAM, and status for every Claude process, auto-refreshed every 1.5 seconds
• System RAM overview — Total, used, and available memory at a glance; all values in GB; processes under 100 MB grouped into "Others" to reduce noise
• Top RAM consumers — See what else is eating your memory (Docker, Postgres, Colima, Xcode, browsers, etc.)
• CPU dashboard — Global CPU usage, core count, and per-process sparkline history over the last 60 samples
• Disk usage — Total/free space, usage gauge, and ~/.claude directory size
• Network I/O — Per-Claude-process network bytes in/out via macOS nettop
• Internet speed — Real-time system-wide upload and download rates
• Adaptive colors — Catppuccin Mocha palette with automatic 256-color fallback for terminals without true color

git clone https://github.com/AgenteMorty/cmon.git
cd cmon
./install.sh

The script will:

1. Install Rust if not present
2. Build the optimized release binary (stripped + LTO)
3. Copy cmon to /usr/local/bin (or a custom path)

cargo build --release
sudo cp target/release/cmon /usr/local/bin/cmon

INSTALL_DIR=~/.local/bin ./install.sh

cmon

• macOS — uses nettop and du for some metrics (not available on Linux)
• Rust toolchain — installed automatically by install.sh if missing

True color: If your terminal supports 24-bit color (iTerm2, Ghostty, Kitty, Alacritty, WezTerm), set COLORTERM=truecolor in your shell profile for the best colors. Without it, cmon uses 256-color approximations that work everywhere.

• Rust — no garbage collector, minimal memory footprint
• ratatui + crossterm — terminal UI
• sysinfo — cross-platform system metrics
• Synchronous event loop (no tokio) — faster compile, smaller binary

This project is licensed under CC BY-NC 4.0 — you can use, modify, and share it freely for non-commercial purposes. Commercial use requires permission from the author.