amux — Agent Multiplexer
Run a parallel AI coding team from your browser or phone
Start new thread

amux - The Agent Control Plane - Run a parallel AI coding team from your browser or phone

by•
amux is an open-source control plane for AI coding agents. Instead of running one Claude Code session at a time, you run 10-50 in parallel — each claiming tasks from a shared kanban board, coordinated so they never duplicate work. Includes a web dashboard, mobile PWA + native iOS app, built-in CRM, Gmail API, browser automation, scheduler, and agent-to-agent messaging. Single Python file. Zero external dependencies. Works with Claude Code, Codex, and Gemini CLI.

Add a comment

Replies

Best
Hunter
📌
Hey PH 👋 — I'm Ethan, I built amux. The origin story: I was running Claude Code to parallelize a big refactor and spending more time babysitting sessions than actually reviewing code. Context would exhaust at 3am, the agent would stall waiting for approval, and I'd wake up to nothing done. So I built a watchdog. Then a dashboard to watch all sessions at once. Then a kanban board so agents could claim tasks without stepping on each other. Then a mobile app so I could steer from my phone. Then email integration so agents could send me updates. One thing led to another. The thing I'm most proud of: the entire server is a single Python file (~30k LOC) with inline HTML/CSS/JS. No build step, no Docker, no Kubernetes. You clone it, run it, and it restarts itself on save. Would love your feedback — especially from anyone running multi-agent setups. What's breaking for you that amux could fix? GitHub: Docs: