wmux — Windows Terminal Multiplexer for AI Agents (cmux alternative)
wmux is LSP-for-terminals — a neutral substrate that lets external tools build workflow intelligence on top of any terminal session.
Native Windows terminal multiplexer with split panes, MCP bridge, and browser automation — purpose-built for running Claude Code, Codex CLI, and Gemini CLI side by side. No WSL required.
Keywords: Windows tmux, tmux for Windows, terminal multiplexer Windows, AI agent terminal, Claude Code Windows, Codex CLI, Gemini CLI, MCP server, Chrome DevTools Protocol, split terminal, multi-agent terminal, browser automation, ConPTY, xterm.js, Electron terminal.
Still running one terminal for your AI coding agents on Windows?
Windows has no native tmux. Without WSL, there was no clean way to run multiple AI coding agents side by side.
wmux fixes this. A native Windows terminal multiplexer + browser automation + MCP server, purpose-built for AI coding agents like Claude Code, Codex CLI, and Gemini CLI. Your AI agent reads the terminal, controls the browser, and works autonomously — all in one window.
Claude Code writes the backend on the left
Codex builds the frontend on the right
Gemini CLI runs tests at the bottom
— all on one screen, simultaneously.
Install in 30 seconds
Winget (recommended — no security warning):
winget install openwong2kim.wmux
Chocolatey:
choco install wmux
Installer: Download wmux Setup.exe
Seeing a "Windows protected your PC" warning? The installer isn't code-signed yet, so Windows SmartScreen flags it as from an unknown publisher. It's safe to proceed — click More info → Run anyway. Installing via winget or Chocolatey above avoids this prompt entirely, since those package managers run in a trusted context.
One-liner (PowerShell):
irm https://raw.githubusercontent.com/openwong2kim/wmux/main/install.ps1 | iex
Why wmux?
1. Your AI agent controls the browser — for real
Tell Claude Code "search Google for this" and it actually does it. wmux's built-in browser connects via Chrome DevTools Protocol (CDP). Click, type, screenshot, execute JS — all done by the AI directly. Works perfectly with React controlled inputs and CJK text (Korean, Japanese, Chinese).
You: "Search for wmux on Google"
Claude: browser_open → browser_snapshot → browser_fill(ref=13, "wmux") → browser_press_key("Enter")
→ Actually searches Google. Done.
2. Multiple terminals in one window
Ctrl+D to split, Ctrl+N for new workspace. Place multiple terminals and browsers in each workspace. Ctrl+click for multiview — see multiple workspaces at once.
ConPTY-based native Windows pseudo-terminal. xterm.js + WebGL hardware-accelerated rendering. 999K lines of scrollback. Terminal content persists even after app restart.
Tmux-style prefix mode — Ctrl+B then a single action key (split, focus, new workspace…) for muscle-memory terminal navigation. 13 actions, fully rebindable.
Floating pane — Ctrl+` for a Quake-style dropdown terminal that lives outside your main layout. Stays alive across toggles.
Scroll bookmarks — Ctrl+M marks the current scroll position, Ctrl+Up/Down jumps between marks. Indicators render on the gutter.
Smart right-click — Windows Terminal style. Selection → instant copy. Empty area → instant paste. Link → small Open / Copy Link menu. Zero modal interrupts.
3. No more asking "is it done yet?"
wmux tells you when your AI agent finishes.
- Task complete → desktop notification + taskbar flash
- Abnormal exit → immediate warning
git push --force, rm -rf, DROP TABLE → dangerous action detection
Not pattern matching — output throughput-based detection. Works with any agent.
4. Automatic Claude Code (MCP) integration
Launch wmux and the MCP server registers automatically. Claude Code just works:
