Introduction

cyber-mux is a small CLI for cross-multiplexer pane control. It gives you one way to drive panes — open, send, submit, read, focus, close, list — regardless of the terminal multiplexer you happen to be inside. It also manages git worktrees bound to a workspace, and can build a whole named template of panes in one call.

Today it supports four backends:

• tmux — the ubiquitous terminal multiplexer (tmux).
• herdr — an agent-aware multiplexer (herdr.dev).
• WezTerm (alpha) — a GUI terminal with a built-in multiplexer (wezterm.org), driven through wezterm cli. Built against wezterm cli --help/the CLI reference rather than a live GUI — treat it as unverified until confirmed against a real WezTerm session.
• Zellij (alpha) — a terminal workspace with a built-in multiplexer (zellij.dev), driven through zellij action. Requires Zellij ≥ 0.44.1. Built from the Zellij docs and CHANGELOG rather than a live binary — treat it as unverified until confirmed against a real Zellij session.

It is kept deliberately narrow: it drives panes and nothing else. No mail, no dispatch, no agent registry — those live in the tools that build on top of it. Those tools can skip the CLI and import the same core directly — see the Library API.

Try it

# What multiplexer am I in?
npx cyber-mux doctor

# Just the backend name (tmux / herdr / wezterm / zellij / none)
npx cyber-mux mode

doctor reports the detected multiplexer, your current pane, and a fast-path pin you can export to skip detection entirely.

Where next

• Multiplexers — the tmux, herdr, WezTerm, and Zellij backends and how their feature sets differ.
• Detection — how cyber-mux figures out where it is running.
• Templates — named, reusable pane pools.
• Worktrees — git worktrees bound to a workspace.
• AXI — the agent-facing output contract every command follows.
• Library API — import the core directly instead of shelling out to the CLI.
• CLI Reference — the full verb surface, one page per command.