Herdr: A Home for Your Coding Agents 🐑

Herdr is a free, open source (Apache 2.0) native app for running coding agents like Claude Code and Codex side by side on macOS, Linux and Windows. Supported agents report whether they are working, done or waiting on you, agents keep running in a local server (herdr-core) when you quit the app, and you customise it by asking your agent to edit its config files.

If you have started running more than one coding agent at a time, you know the routine. One tab has Claude Code refactoring a module, another has Codex writing tests, a third is chewing on a migration. Then you spend half your day cycling through tabs asking the same question: is anyone waiting on me? Herdr is built for exactly that problem.

"Made for your agents. Remade by them."

🐑 What Is Herdr?

Herdr runs real terminals, organised around how agent work actually happens:

• Spaces are for repos.
• Tabs are for tasks.
• Each agent reports whether it is working, done, or waiting, and the sidebar and tabs surface whoever needs your attention.

That last point is the whole game. Instead of polling every terminal, you glance at the sidebar and go where you are needed.

The source is split across three repos under the herdrdev GitHub organisation: herdr-app, herdr-core and the original herdr TUI.

🔥 Features for Running Multiple Coding Agents

Agents Keep Running

Agents do not live inside the app window. They run in herdr-core, a local server. Quit the app, update it, or crash it, and your terminals keep going.

One caveat worth knowing: if the machine reboots or the server stops, panes come back with a shell in the same folder, but Herdr does not restart or resume the agents. You start them again yourself.

Multiple Machines, One Sidebar

Any machine you can reach over SSH (a VPS, a VM, a cloud sandbox) can join the same sidebar:

hc machine add vps

Agents on that machine show up in the same list as your local ones. If you offload heavy agent runs to a beefy remote box, this is a big quality-of-life win.

Agents That Coordinate

This is the part I find most interesting. One agent can start helper agents, steer them, and receive their reports, all through the hc CLI:

# Start a background reviewer that reports back to tab t1 when done
hc run --bg codex "review my diff. Done: hc send t1"

# Nudge a running helper
hc send t3 "take another look"

Reports come back tagged with their source, like <hc from="t3 reviewer">. So your main Claude Code session can spin up a Codex reviewer, keep working, and pick up the review when it lands. That is a small, practical form of multi-agent orchestration without any framework. It is also the easiest way I have seen to run multiple Claude Code agents that actually talk to each other.

Works With Any Agent

Because Herdr runs real terminals, anything that runs in a shell runs in Herdr. For status reporting, agents use OSC 7501, an open escape-sequence protocol. Any program can report its own state with a single printf:

printf '\e]7501;state=blocked:app=my-agent\e\\'

That means your own scripts and homegrown agents can light up the sidebar exactly like Claude Code or Codex do.

🛠️ Customise It by Asking Your Agent

Everything about Herdr's look and behaviour lives in plain files:

~/.config/herdr-app/
├── AGENTS.md
├── settings.json
├── keys.json
├── layouts/
├── mods/
└── themes/

The AGENTS.md file explains all of it to your agent. So instead of reading docs and editing JSON yourself, you tell your agent what you want:

"put a timer on each tab: how long its agent has been working."

and it builds it. Layouts rearrange how your agents are shown (for example, a board with "needs you", "working" and "done" columns, switchable with ctrl-shift-l). Mods extend behaviour, like a notification mod that pings ntfy.sh when an agent needs you.

There is also a community gallery. At the time of writing it lists 215 layouts, 435 mods and 393 contributors, and you can add someone else's layout or mod straight from its herdr.dev/layouts/... or herdr.dev/mods/... address.

🚀 Getting Started

Herdr runs on macOS (Apple silicon), Linux (x86_64 and arm64) and Windows (x86_64). Remote machines join over SSH regardless of their OS.

On macOS or Linux:

curl -fsSL https://releases.herdr.dev/app/install.sh | sh

On Windows (PowerShell):

powershell -c "irm https://releases.herdr.dev/app/install.ps1 | iex"

Mac users can also grab the .dmg directly. As always, read an install script before piping it into your shell.

Then:

1. Start an agent from the + in the tab bar (Claude Code, Codex, or any program).

2. Point it at the guide:

   Read ~/.config/herdr-app/AGENTS.md
   

3. Ask it for whatever you want Herdr to do.

⚠️ A Few Gotchas

• Herdr TUI agents don't show up. Agents started in the original TUI will not appear in the app (see the FAQ below).
• Don't nest in tmux. If an agent is not showing up, update it and run it directly in a Herdr pane rather than inside tmux.

❓ FAQ

Is Herdr free?

Yes. The app, herdr-core and the Herdr TUI are all open source under the Apache 2.0 license.

Does Herdr work with Codex and other agents?

Yes. Herdr runs real terminals, so any agent works. Agents that support OSC 7501, an open status protocol, also report whether they are working, done or waiting.

Does Herdr resume agents after a reboot?

No. Panes come back with a shell in the same folder, but you start the agents again yourself.

What is the difference between Herdr and Herdr TUI?

Herdr TUI is the original terminal app with its own server. The Herdr app is separate, and the two do not share agents.

🤔 Final Thoughts

I've written before about my terminal setup and about using screen to detach and reattach sessions. Herdr feels like the natural next step for the agent era: a persistent session manager that knows what is running in each terminal and whether it needs you.

The "remade by agents" idea could easily have been a gimmick, but putting all configuration in plain files with an AGENTS.md guide is a genuinely sensible design. Your agent already knows how to edit files; Herdr just gives it a map.

If you are juggling more than a couple of agents a day, it is free, open source, and worth an afternoon. Grab it at herdr.dev.

Happy herding! 🐑