Ghostty-based macOS terminal with vertical tabs and notifications for AI coding agents
https://cmux.com
`surfaceConfig.env_vars` are overrides layered on top of Ghostty's default subprocess environment. Starting the env map from `ProcessInfo.processInfo.environment` re-introduced variables that Ghostty intentionally strips (e.g. LANGUAGE, Xcode DYLD vars), bypassing its env filtering. Start with an empty map and add `getenv()` fallbacks for PATH, SHELL, ZDOTDIR, and GHOSTTY_RESOURCES_DIR so critical values are still resolved without leaking the full app environment. |
||
|---|---|---|
| .claude/commands | ||
| .github/workflows | ||
| Assets.xcassets | ||
| CLI | ||
| docs | ||
| docs-site | ||
| ghostty@4713b7e236 | ||
| GhosttyTabs.xcodeproj | ||
| GhosttyTabsUITests | ||
| homebrew-cmuxterm@c12d753824 | ||
| node_modules | ||
| Resources | ||
| scripts | ||
| Sources | ||
| tests | ||
| .gitignore | ||
| .gitmodules | ||
| .vercelignore | ||
| AGENTS.md | ||
| bun.lock | ||
| CHANGELOG.md | ||
| CLAUDE.md | ||
| cmux-Bridging-Header.h | ||
| CONTRIBUTING.md | ||
| ghostty.h | ||
| package.json | ||
| Package.resolved | ||
| Package.swift | ||
| README.md | ||
| TODO.md | ||
cmuxterm
A lightweight native macOS terminal with vertical tabs and notifications for AI coding agents
Features
- Native macOS app — Built with Swift and AppKit, not Electron. Fast startup, low memory.
- Vertical tabs — See all your terminals at a glance in a sidebar
- Notification panel — Tabs flash when AI agents (Claude Code, Codex) need your attention
- Lightweight — Small binary, minimal resource footprint. No bundled browser engine.
- GPU-accelerated — Powered by libghostty for smooth rendering
Install
Homebrew:
brew tap manaflow-ai/cmuxterm
brew install --cask cmuxterm
Or download the DMG directly.
Why cmuxterm?
Running multiple AI coding agents? cmuxterm helps you manage them. Instead of losing track of which terminal needs input, the notification panel shows you exactly where to look.
A native macOS app means it launches instantly, uses minimal RAM, and feels right at home on your Mac.