amux CLI
A control plane that runs Claude Code, Codex and Gemini CLI agents in parallel with a shared task board and schedules
- 544 ★ on the repository
- Last commit
Broad access
Why this level
- Agents run autonomously, including overnight, with access to working directories and commands on your computer
- Installing through `curl ... | bash` downloads code and, with your consent, installs Rust and tmux
- Agent actions are approved from the dashboard and a phone, so access to the dashboard must be protected
Install
- Works in
Manual install
curl -fsSL https://amux.io/install.sh | bashClones amux into ~/Dev/amux and asks before installing Rust and tmux. Needs tmux 3.2 or newer and at least one agent.
Access and checks
How we reviewBroad accessThe tool can change data, run commands or handle secrets.
- Agents run autonomously, including overnight, with access to working directories and commands on your computer
- Installing through `curl ... | bash` downloads code and, with your consent, installs Rust and tmux
- Agent actions are approved from the dashboard and a phone, so access to the dashboard must be protected
This is third-party code. Review the repository files before installing.
What it does
amux runs several AI coding agents in parallel from one dashboard or from a phone: Claude Code, Codex and Gemini CLI, plus OpenCode and Ollama. Each agent works in its own tmux session with a durable identity. It is a single Rust binary with SQLite storage on your own machine, and the web dashboard opens at `https://localhost:8824`. Agents share a task board where claiming a card is atomic, so two agents never take the same work. They message each other, and the messages arrive at turn boundaries. Crashed or stuck sessions restart and pick up where they stopped. There are schedules such as `daily at 9am` or `every 15m`, overnight loops, and groups of agents with shared memory and layered environment variables. The model can be switched per agent. The `amux-rs` command-line client manages the board, agents, messages and schedules. You can watch, steer and approve from a Mac app, an iPhone and iPad app, or a browser on Android. For round-the-clock work the authors offer a paid hosted server, amux Cloud, at `yourname.amux.io`.
Who it is for. For developers who run several agents on different tasks and want a shared board, schedules and control from a phone.
Good fit when
- Several agents should pick tasks from a shared board without duplicating work
- You need overnight or recurring scheduled runs, for example every 15 minutes
- You want to watch agents and approve their actions from a phone
- Different agents need different models, including local ones through Ollama
- Agents crash or hang and you need an automatic restart that resumes the work
Not a fit when
- You work with one agent in one window: a plain terminal is enough
- You are on Windows: amux runs on macOS and Linux
- You want to resell amux commercially: the MIT with Commons Clause license requires a separate license for that
Example request
Create three payment-module refactoring tasks on the board, give two to two Claude Code agents and put the third, with tests, on an overnight schedule.Limitations
It runs on macOS and Linux and needs tmux 3.2 or newer and at least one agent (Claude Code, Codex or Gemini CLI). The installer clones amux into `~/Dev/amux`, data lives in `~/.amux`, and Rust is installed through rustup with your consent.
The license is MIT with Commons Clause: you may use, modify and self-host it, while commercial resale needs a separate license. The Mac app is not notarized yet, so macOS blocks it on first launch and you must choose Open Anyway in privacy settings. The dashboard uses a self-signed HTTPS certificate.
The paid amux Cloud is run by the authors and needs your own Claude, ChatGPT or Gemini account or an API key.
How to disable. On Linux turn the services off with `systemctl --user disable --now amux-server amux-builder.timer`; on macOS stop the launchd agents the installer created. Then delete `~/Dev/amux` and the data in `~/.amux`.
FAQ
How do I install amux?
The `curl -fsSL https://amux.io/install.sh | bash` script installs amux in one command. To read the code first, clone the repository and run `./install.sh`: it is the same installer.
Which agents does amux support?
Claude Code, Codex, Gemini CLI, OpenCode and Ollama. Each agent runs in its own tmux session, which you can open directly in a terminal.
How do two agents avoid taking the same task?
Claiming a card on the shared board is atomic, so two agents never get the same work. Agents also message each other and hand work off without your involvement.
What can I do from a phone?
Watch agents, steer them and approve their actions. There is an App Store app for iPhone and iPad, and on Android the dashboard installs from the browser with Install app or Add to Home Screen.
What are .mdai files?
They are computed markdown files whose content a model assembles from linked files. The format is described in the amux docs.
Discussion
No questions or reviews yet.
Related
CLI toolsOpen-source personal AI assistant on your own machine: answers in Telegram, Slack, Discord and WhatsApp, extended with skills and plugins
An engineering system for coding agents: 68 agents, 292 skills, hooks, memory, rules and the AgentShield security scanner
DeepSeek Harness is an open-source AI agent from DeepSeek where every feature is a plugin. It runs in the browser, on macOS and on Windows.
A self-improving agent from Nous Research with a TUI, messaging gateway, cron jobs and skills it writes itself
An open source coding agent for the terminal and desktop with build and plan modes
Ollama runs open models on your own machine and serves them to agents through a local API compatible with OpenAI and Anthropic.
Comparison with similar tools
Entries that share the main search query, and how they differ.
| Entry | Official | Repository stars | Main difference |
|---|---|---|---|
| amux (this entry) | no | 544 | |
| Claude Squad | no | 8,656 | a terminal app for several agent sessions, without a shared board, schedules or phone apps |
| Hive | no | 579 | a local browser workbench where agents work as a team under an orchestrator |
| Octogent | no | 1,413 | a local dashboard for many Claude Code sessions with child agents |
README in short
It is built on eight primitives: board, agents, schedulers, filesystem, groups, memories, environment and messages, and new capabilities are composed from them. The server is written in Rust, and the web dashboard is embedded in the same binary.