My terminal is still my IDE, but the way I work has become increasingly agent-centric. Herdr gives me the keyboard-driven navigation I loved in tmux while making it easy to move between agents, see which ones need attention, and check in from my phone when I step away.

Switching from Tmux to Herdr

I’ve written before about why my terminal is my IDE.

My development environment revolves around Neovim, the command line, and tmux. I like working in a keyboard-driven environment where everything is fast, composable, and available over SSH. That workflow has served me well for years.

But the way I build software is changing.

I’m no longer working in only one editor with a few supporting terminal windows. Increasingly, I have several coding agents working alongside me—sometimes across multiple repositories, branches, or git worktrees. One agent may be implementing a feature while another investigates a bug and a third runs tests or reviews a change.

Tmux can run all of those processes, but it doesn’t really understand what they are. To tmux, an agent is just another shell in another pane.

That is why I recently switched to Herdr.

Herdr is an agent-centric terminal multiplexer. The easiest way to describe it is that it brings the tmux model into an environment where coding agents are first-class participants. It preserves the terminal-native workflow I already like while making it much easier to manage a growing collection of agents.

Three benefits have made the switch worthwhile.

1. Agents Feel Like Tmux Sessions

I have years of muscle memory built around tmux. I can create sessions, move between them, split panes, detach from a machine, and reconnect later without thinking much about it.

Herdr keeps that keyboard-driven style of interaction. It organizes work into what it calls spaces, and I can quickly move between spaces and the agents running inside them.

But the important part isn’t simply that Herdr has keyboard shortcuts.

Herdr lets me move between agents the way I used to move between tmux sessions.

That distinction matters.

When I have several agents running, I don’t want to hunt through windows and panes trying to remember where each one lives. I want agents to be navigable objects in their own right. I want to jump directly from the agent working on authentication to the one debugging a deployment and then to the one waiting for feedback on a test failure.

Herdr makes that movement natural.

I’ll say it again because this is the feature that fundamentally changed the workflow for me:

Herdr allows me to move between agents like they’re sessions in tmux.

When you’re juggling a bunch of agents, that’s incredibly important.

2. My Agent Workflow Is No Longer Tied to My Desk

The other major benefit is mobile access.

I use Herdr with Moshi, an iPhone terminal app designed with persistent terminals and coding agents in mind. Together, they give me a clean way to check on my agents when I step away from the computer.

I can pull out my phone, see which agents are still working, notice which one has stopped, and provide whatever input it needs.

This does not mean I’m trying to replace my desktop development environment with an iPhone. I’m not planning to spend an afternoon writing code on a tiny touchscreen.

The phone is a control surface.

An agent may work independently for several minutes and then stop because it needs permission, clarification, or a decision. Without mobile access, that agent sits idle until I return to my computer. With Herdr and Moshi, I can respond while I’m elsewhere in the house, waiting in line, or grabbing coffee.

The feedback loop stays alive even when I’m not sitting at my desk.

That changes how I use agents. I’m more comfortable starting longer-running tasks because I know I don’t have to remain physically attached to the terminal just to keep them moving.

3. Herdr Shows Me Where My Attention Is Needed

The third benefit is visibility.

A traditional terminal multiplexer shows me processes. Herdr gives me a higher-level view of the agents themselves, including whether they are working, blocked, idle, or finished.

That becomes increasingly valuable as the number of agents grows.

With one or two agents, it’s easy to remember what each one is doing. With six or eight, the limiting factor is no longer terminal space. It’s attention.

Which agent is still making progress?

Which one completed its task?

Which one is waiting for me?

Which one has been sitting at a prompt for the last ten minutes?

Herdr helps answer those questions without requiring me to inspect every pane individually. Instead of repeatedly cycling through terminals, I can focus on the agents that actually need intervention.

That makes the workflow feel less like managing a pile of shell processes and more like supervising a small team.

From a Terminal-Centric Workflow to an Agent-Centric One

I still believe the terminal is my IDE.

I still use Neovim. I still want real shells, persistent sessions, keyboard shortcuts, SSH access, and tools that compose cleanly. Herdr doesn’t ask me to abandon that environment for a separate graphical agent dashboard.

Instead, it adapts the terminal workflow to the way I’m now building software.

Tmux was designed around terminals and processes. Herdr is designed around terminals, processes, and agents.

That gives me three things I increasingly need:

Fast keyboard navigation between agents
Mobile access and notifications when I step away
A clear view of which agents need my attention

Those benefits alone make Herdr worth using.

The biggest change, though, is not any individual feature. It’s that my agent workflow is no longer confined to the chair in front of my computer.

I can start several agents, step away, receive a notification when one stops, check its progress from my phone, give it direction, and let it continue working. When I return to my desk, the same real terminal sessions are waiting for me.

My terminal is still my IDE.

Now it also has a place to manage the herd.