netmon

A terminal network speed monitor for Linux and macOS. Shows live download/upload throughput as a scrolling graph, viewable over several timeframes.

netmon  iface: wlan0 (1/3)  window: 5m  down: 1.2 MB/s   up: 84 KB/s

Download   peak 3.4 MB/s  avg 980 KB/s
3.9 MB/s
                                                ▁▂▃▅▆▇████▇▆▅▄▃▂▁▂▃▅▇█
0

Upload   peak 210 KB/s  avg 76 KB/s
240 KB/s
                                                    ▁▂▃▃▄▄▃▂▁▁▂▃▄▄▃▂▁
0
        -5m                                                     now
[1]60s [2]5m [3]15m [4]1h [5]6h [6]24h
n: next iface   p: prev iface   q: quit

Features

• Live download/upload graphs, sampled once per second from the kernel's per-interface byte counters (/proc/net/dev on Linux, sysctl(NET_RT_IFLIST2) on macOS)
• Switch between six timeframes (60s, 5m, 15m, 1h, 6h, 24h) without losing history — up to 24h of per-second samples are kept per interface
• Cycle through all detected network interfaces
• Auto-scaling y-axis, peak/average stats for the visible window

Requirements

• Linux or macOS
• A C++17 compiler (g++ or clang++)
• ncurses development headers — the wide-character build (ncursesw) where one exists; macOS's system libncurses is already wide-capable

On Debian/Ubuntu:

sudo apt install build-essential libncursesw5-dev

On macOS, the Command Line Tools are enough (xcode-select --install). Homebrew's ncurses is used automatically if installed.

Build

make

This produces a netmon binary in the project root.

Run

./netmon            # auto-selects the first active non-loopback interface
./netmon wlan0       # start on a specific interface

Keybindings

Key	Action
1–6	Select timeframe (60s/5m/15m/1h/6h/24h)
n / Right	Next interface
p / Left	Previous interface
q	Quit

Project layout

src/main.cpp     - ncurses UI, input handling, sampling loop
src/netstat.h/.cpp - per-interface byte counters (Linux + macOS backends)
src/model.h      - per-interface history ring buffer, downsampling, formatting
Makefile

Disclaimer

This project was written with the assistance of AI. The code and documentation were largely AI-generated and then reviewed by a human. Review it yourself before relying on it for anything important.