---
title: "herdr"
category: "AI"
tagline: "A terminal workspace for running several AI coding agents at once, and coming back to them later."
source: "AUR"
package: "herdr"
command: "herdr"
website: "https://herdr.dev/"
code: "https://github.com/herdrdev/herdr"
links:
  - name: "herdr quick start"
    url: "https://herdr.dev/docs/quick-start/"
---

## What it is

herdr is a terminal workspace manager built for AI coding agents. It splits your terminal into panes, like tmux, and runs your agents in them: Claude Code, opencode and others. Each pane is marked working, blocked or idle, so you can see at a glance which agent is waiting for an answer.

Its terminals keep running in a background server when you close the window or lose an SSH connection; run `herdr` again to pick up where you left off. It can also gather sessions from several machines into one window.

## Why it's here

- **Many agents, one view.** Run agents across projects side by side and spot the one that's stuck.
- **Walk away.** Detach with `ctrl+b q` and the agents keep working.
- **Uses what you have.** It doesn't replace your agents or your terminal; it runs them.

## On CoreDesktop

herdr is installed from the AUR. Start it in a project directory:

```
herdr
```

Agents you start by name in its panes, such as `opencode` or `hermes`, still run in CoreDesktop's [AI sandbox](/ai-sandbox/). `update` keeps herdr current with the rest of the AUR packages.