Verified project record
manaflow-ai/cmux
cmux is a native macOS terminal built on Ghostty that helps developers monitor parallel AI coding agents with vertical tabs, notification rings, and a scriptable CLI/socket API. It targets the problem of losing track of which agent needs attention when many agents run at once.
Project overview
It addresses a concrete operational pain point—unreadable tabs and context-free terminal notifications when running many agents in parallel—with agent attention rings, a jump-to-unread panel, and full programmatic control over panes and an in-app browser.
- Project type
- AI Coding
- Use cases
- Coding & Development · Automation
- Deployment
- Refer to project documentation
- License
- License pending
Best for
- macOS developers who run multiple AI coding agents in parallel and need to see which agent needs attention, and who want terminal-plus-browser workflows controllable from a CLI or socket.
- Developers who already use Ghostty and want their existing config for themes, fonts, and colors to carry over.
- Claude Code users who want to run teammate mode with a single command (cmux claude-teams).
Key capabilities
- Panes get a blue ring and tabs light up when coding agents need attention. Notifications are triggered by terminal sequences OSC 9/99/777 or a cmux CLI hook, with a panel and jump-to-unread navigation.
- A real browser pane can be split next to the terminal with a scriptable API: navigate, snapshot the DOM/accessibility tree, click, type, evaluate JavaScript, and read console and network activity.
- Every action is available through the cmux CLI and a Unix socket: create workspaces, split panes, send input, read screen contents, take screenshots, and drive the browser.
Limitations and risks
- cmux is a native Swift + AppKit application and is available on macOS only for now; users on other platforms are not served.
- GPU requirements, telemetry practices, and an end-to-end data boundary are not documented, so no conclusion can be drawn about where data flows during operation. cmux itself is documented as requiring no paid service, but any coding agents you run through cmux may themselves require remote models, APIs, or paid services; those dependencies are separate from cmux's local orchestration.
Getting started
- Download the DMG from GitHub releases, open the .dmg and drag cmux to the Applications folder, then launch cmux. A Homebrew cask is an alternative. No coding is required to get started.
Alternatives and comparisons
- A self-hosted developer control center for running coding agents across local, Docker, VM, or cloud backends; relevant when multi-backend agent execution matters more than a native macOS terminal UI.
- A unified multi-provider LLM API and agent runtime with an interactive coding agent CLI in the terminal; an option when you want an agent runtime/CLI rather than a monitoring terminal.
Project comparisons
Evidence and sources
- README: A Ghostty-based macOS terminal with vertical tabs and notifications for AI coding agents
- README: Claude Code's notification body is always just "Claude is waiting for your input" with no context, and with enough tabs open I couldn't even read the titles anymore.
- README: Panes get a blue ring and tabs light up when coding agents need your attention
- README: Split a browser alongside your terminal with a scriptable API ported from
- README: Open the `.dmg` and drag cmux to your Applications folder. cmux auto-updates via Sparkle, so you only need to download once.
AI Search
Find projects, verify facts, compare options, or turn a complex need into an actionable plan
Try a searchA click only fills the search box; you stay in control
Project Details
0