wmux

What wmux is

wmux is a Windows terminal multiplexer made for coding agents. The README calls it the original visibility layer for coding agents on Windows, with the goal that you never hunt for the one agent that is waiting on you. It is built on Electron and xterm.js and takes inspiration from cmux, while its agent-visibility model is modelled on herdr. The core problem it solves is simple to state: when you run several agents across many panes, you lose track of which one is blocked and which one needs attention. wmux collects those agents into one place so you can see status at a glance. Split panes and tab-like surfaces are part of the interface, so the layout feels like a conventional terminal multiplexer with an agent roster added on top. Because it is Windows only, the download and release badges point to Windows installers, and the project is positioned as the native option for that platform rather than a cross-os tool. The visible roster is the feature that separates wmux from a plain terminal splitter, because it turns passive panes into queryable agents. For someone running a handful of assistants at once, that single pane of truth is the reason the tool exists at all.

What changed in 2.0

Version 2.0 broadens who wmux can see. Before this release, wmux could only report an agent's state if that agent told it, which covered Claude Code through hooks, OpenCode through a plugin, and Kiro. Codex, Gemini, Aider, Amp, Cursor, and Copilot ran in panes wmux could display but could not read. The 2.0 release closes that gap with three layers, each filling what the one above left open. The first layer is one roster for every agent in the window, with a banner over the workspace list answering who needs me across all workspaces, ranked by who has waited longest. The second layer is a keyboard shortcut, Ctrl+Shift+A, that opens an agent navigator listing every agent with blocked ones first, filterable by keys and jumpable with the arrow and enter keys. This is the change that makes wmux useful beyond a single agent brand. The third layer is left to fill remaining gaps for agents that still do not report, so coverage keeps improving. The README is clear that each layer only covers what the previous one missed, which is a sensible way to grow visibility without a single hard dependency on any one agent's hook system. The result is broader coverage without waiting for every agent to add a hook.

Using it day to day

The README describes features such as workspaces, surfaces that act as tabs, split panes, and a browser view, with a keyboard shortcuts section explaining each. The agent navigator is the headline interaction: you press the shortcut, see blocked agents first, filter to what matters, and jump straight to the pane that needs you. For someone running several coding assistants at once, that saves the repeated manual scanning of terminal windows. The project is MIT licensed and built on Electron plus xterm.js, so the surface is a desktop app rather than a pure command-line tool. The inspiration credits and the modelled visibility approach are acknowledged in the README, which is a fair note about where the design came from. If your work happens in Windows terminals with more than one agent, wmux turns scattered panes into one manageable queue. The blocked-first ordering in the navigator is the detail that makes it practical, because it answers the only question that matters when you have many agents running: which one is stuck waiting on me right now, and how long has it been there. Everything else in the UI supports that single piece of triage. For heavy users the navigator alone justifies running wmux over a plain splitter.

Editorial conclusion

wmux is published under the MIT license, built on Electron and xterm.js, and targets Windows only.