July 30, 2026
Tmux, tantrums, and bot babysitting
Agent-Manager: A Tmux TUI for Running Claude Code, Codex and OpenCode
One screen to boss around your AI helpers — and the comments are already arguing
TLDR: Agent-Manager puts several AI coding assistants in one simple control screen so users can track them without hopping between windows. Commenters liked the problem it solves, but the thread quickly turned into a "haven’t we seen this already?" showdown with rival tools stealing some spotlight.
A new app called Agent-Manager is basically trying to become the air-traffic controller for people who unleash multiple AI coding bots at once. Instead of juggling a chaotic pile of terminal windows, users get one screen showing which bot is working, which one is waiting, and which one has quietly face-planted. It can even send a message to an AI helper without opening its window, revive a dead session, and show file changes in a readable review view. In plain English: it’s a control room for your robot coders.
But the real action is in the reactions. Creator yoanwaidev pitched it with a painfully relatable complaint: the worst part isn’t the coding, it’s discovering one bot has been stuck forever on a permission pop-up while you were busy checking twelve other tabs. That struck a nerve. One commenter simply dropped a deadpan “Nice”, which honestly has big internet-shrug energy. Others came in swinging with the classic tech-community question: wait, isn’t this just like agent-deck? And then another name got tossed into the ring — Superlogical — turning the thread into a mini cage match over who really owns the “many AIs, one dashboard” idea.
The funniest twist? One self-described old-school developer admitted tools like this completely changed his setup after trying a rival product, basically giving Agent-Manager a compliment wrapped in a betrayal. So the vibe is clear: people like the idea, but the comments are serving the usual spicy combo of “cool tool,” “seen it before,” and “my favorite competitor already won me over.”
Key Points
- •Agent Manager runs multiple AI coding agents in parallel from one terminal, with each agent isolated in its own tmux session.
- •The interface provides a live status list grouped into a project tree and allows prompts to be sent directly to sessions without attaching.
- •Built-in status detection supports Claude Code, OpenCode, Codex, and Grok Build, while other CLI tools can be added through configuration.
- •The tool includes review features such as full-file syntax-highlighted diffs and line comments that are sent back to the agent as prompts.
- •Installation is available via Homebrew, Go, prebuilt binaries, and WSL2 for Windows users, and sessions persist after the manager quits.