Pertmux - A TUI to Unify Coding Agents, MRs and Worktrees
Pertmux - A TUI to Unify Coding Agents, MRs and Worktrees
Running multiple AI coding agents at once is becoming standard practice. The problem is not running them - it is keeping track of what each one is doing. You have 3-5 agents working in parallel across different worktrees, each generating merge requests, and you are Alt-Tabbing between terminal windows trying to remember which pane has which task.
The Context Switching Tax
Every time you switch between agent panes, you lose context. You forget what Agent 2 was working on. You confuse the MR from Agent 3 with the one from Agent 1. The productivity gain from parallel agents gets eaten by the overhead of managing them.
This is not a tmux problem. Tmux gives you panes but no semantic understanding of what is happening in each one. You need a layer on top that knows which agent is tied to which worktree, which MR, and what the current status is.
What a Unified TUI Needs
A tool like Pertmux should give you:
- Agent-aware panes - each pane knows which coding agent is running and what task it is assigned to
- Worktree mapping - visual indicator of which git worktree each agent is operating in
- MR status integration - see merge request state without leaving the terminal
- Cross-pane search - find output across all agents at once
- Priority flagging - highlight which agent needs your attention right now
The key insight is that the unit of work is not the terminal pane. It is the agent-worktree-MR triple. Your TUI should organize around that unit.
Why This Matters for Desktop Agents
Desktop AI agents face the same challenge at the application level. When an agent is operating across multiple apps simultaneously, you need a unified view of what it is doing and why. The same principles that make Pertmux useful for coding agents apply to any multi-context AI workflow.
Fazm is an open source macOS AI agent. Open source on GitHub.