Claude Code multi-agent

Omux

Orchestrate AI coding agents (Claude Code, Codex) as parallel subagents over tmux — a loop-engineering runtime with auto-continue, execute-then-review, and cross-session memory.

Install

Install
npm install -g @happenmass/omux

Found in the project README. Check it for setup steps after installing.

From source
git clone https://github.com/Happenmass/omux.git

From the README

omux

Run your CLI coding agent unattended, in parallel, at scale.

omux is a meta-agent that orchestrates the CLI coding agent of your choice — Claude Code, Codex, or anything else — through tmux. It spawns multiple instances, handles state and confirmations, remembers across sessions, and lets you walk away.

Website · Demo video · How it works · Install · FAQ

The problem

Claude Code and Codex are honestly some of the best products of this era. But if you use them daily, you know this moment.

You sit down. Fingers on the keyboard. You think for a while… and then you type implement this feature @somefile.md and hit enter.

You know better. The prompt *should* carry the background, the boundaries, the tasks broken out, a clear definition of done — maybe a test-first plan. You send the one-liner anyway.

I used to think that was a discipline problem. It isn't — it's a bandwidth problem. The intent in your head is high-dimensional; a one-line prompt is a lossy compression of it. Working memory holds a handful of chunks, you type at conversational speed, and you're juggling three projects at once. The bits you drop don't vanish — the model quietly fills them from its own prior, bakes those guesses into the code, and hands them back at review time, with interest. In a single chat turn that's a small loss. In an hours-long agent run, a wrong assumption at step 1 conditions every step after it.

So we compensate with harnesses — hooks, MCP servers, skills, elaborate CLAUDE.md files. It works, but you're paying for human limits with complexity, and it's a power-user game.

And that's before the operational tax. You still can't walk away: the agent pauses on every destructive action to make you confirm. You still run one at a time per task — want one agent on the backend…
Read the full README on GitHub

Details

Language
TypeScript
Last updated
14 July 2026
Topics
agent-orchestratoragentic-workflowai-agentsautonomous-agentsclaude-codeclicodexcoding-agentllm-orchestrationmcpmulti-agenttmux

FAQ

What is Omux?

Orchestrate AI coding agents (Claude Code, Codex) as parallel subagents over tmux — a loop-engineering runtime with auto-continue, execute-then-review, and cross-session memory. It is a multi-agent for Claude Code, from Happenmass/omux on GitHub.

How do I install Omux?

Run npm install -g @happenmass/omux, then follow the setup in the README.

Is Omux popular?

Happenmass/omux has 101 GitHub stars and 3 forks. It was last updated on 14 July 2026.

More multi-agent & worktrees

Browse the most popular Claude Code mods, all multi-agent & worktrees, or the skills and MCP server directories.