Herdr on Windows: Install It, Add Mods and Track Agent Usage

Herdr is a free, open-source desktop app for running AI coding agents like Claude Code and Codex in one window. It also runs on Windows, which is where I tested it for this guide. If you have tried to manage several coding agents from plain PowerShell windows, you know how quickly it turns into a mess of tabs you cannot tell apart.

This guide covers the Windows side of Herdr: installing it from PowerShell, adding community mods with a single prompt, tracking your Claude Code and Codex usage, cleaning up layouts you no longer use, and using Herdr to build a small Windows app of your own.

Prefer to watch? See the full walkthrough on YouTube. I build the whole thing on screen.

Product details below were checked against Herdr’s official site and its gallery on November 10, 2026 (ICT). Herdr changes quickly, so treat version-specific details as a snapshot. If you want the general desktop setup, including workspace migration and choosing a first layout, read the Herdr desktop app guide. For the original terminal version, see running Claude Code, Codex and OpenCode in one terminal with Herdr.

What Herdr gives you on Windows

Herdr is a native app for running coding agents. The agents themselves run inside herdr-core, a small background server on your machine, so closing or updating the app window does not stop them. Each agent reports whether it is working, done or waiting on you, and the sidebar shows which one needs attention.

According to the official FAQ, the app supports Windows on x86_64, plus macOS on Apple silicon and Linux on x86_64 and arm64. The app, herdr-core and Herdr TUI are open source under the Apache 2.0 license, and the app is free. You still need your own access to whichever agents you run inside it, such as a Claude or ChatGPT plan or API keys.

The part that makes Herdr different is how you customize it. Its layout, tabs, keys and themes live as plain files, and an AGENTS.md file in the config folder explains them to your agent. You describe the change you want and the agent edits the files.

Step 1: Install Herdr from PowerShell

Open the Herdr download page and copy the Windows command. At the time of writing it is a one-line PowerShell installer:

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

Paste it into a PowerShell window and run it. The installer adds a Start menu entry and sets up herdr-core if you do not already have it. After that, the app updates itself. The same line also upgrades an existing install.

In my recording, the one-liner did not work at that moment, and I used the copy that was already installed. If the same happens to you, wait a little and try again, or check the download page for changes before troubleshooting your machine. As with any command that downloads and runs a script, read it first if you are unsure what it does.

Step 2: Give each project its own tab

Once Herdr is open, start an agent from the + in the tab bar. You can pick Claude Code, Codex or any other program that runs in a terminal. Herdr’s own guide recommends a space for each repo and a tab for each task.

In the video I have several projects open at once, each running in its own tab, with Claude Code and Codex side by side. That is the main practical gain over PowerShell. You can see every agent at a glance, and the status indicator tells you which one is blocked and waiting for an answer.

Step 3: Point your agent at the AGENTS.md guide

Before you ask Herdr for any change, have your agent read its guide. The official docs suggest starting every customization with this prompt:

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

That file tells the agent what it is allowed to change: settings, keys, layouts, mods and themes. Starting here saves you from an agent guessing at file locations or editing the wrong thing. On my Windows install, pasting the prompt exactly as written worked fine.

Step 4: Install a mod from the gallery with one prompt

The Herdr gallery lists community layouts and mods. Layouts change the whole window. Mods add features, such as a usage panel or a different way to switch focus between panes. Each gallery entry has a one-line prompt you copy and paste into Herdr, plus a link to its GitHub repository.

In the video I install a mod called Insights. Its gallery prompt looks like this:

Read ~/.config/herdr-app/AGENTS.md, then add this mod: herdr.dev/mods/GaganSD/coding-agent-toolkit--herdr-insights

The agent reads the guide, pulls the mod’s files from GitHub into a fresh folder, and adds them next to your existing setup. When it finished, I simply asked how to use the mod, and the agent explained it. In my install, Insights added a new layout with an Insights button beside the tabs.

Per its gallery description, Insights reads the session data that coding agents already leave on your disk and shows token usage for tools including Claude Code, Codex, OpenCode and Pi. In the video it gives me a day-by-day and week-by-week breakdown, lets me switch providers or show all of them, and splits usage by model.

If you want structured help setting up agent workflows like this one, AI Income Lab has courses and a community where you can ask questions while you build.

Track Claude Code and Codex limits from the status bar

Insights is one of many usage mods in the gallery. I also built my own, listed as Herdr Agent Usage, which puts Claude Code and Codex plan usage in the status bar. I use it to see how much of the five-hour window and the weekly limit I have left without opening a separate dashboard. Plan limits are set by Anthropic and OpenAI and change from time to time, so treat any mod’s reading as a convenience and check your account page when the numbers matter.

Step 5: Build your own mod and share it

Creating my usage mod was mostly a matter of telling Herdr what I wanted and letting the agent modify the config files. It did not take long. I also built a second mod, Herdr Auto Focus, which moves focus to whichever pane the mouse is over so you do not have to click each time.

To share a mod on the official gallery, you put it on GitHub with a preview image and a description. The submission details are in the Herdr docs. The simplest route is to ask Claude Code, or whichever agent you use, to read those docs, prepare the repository for your mod and upload it to GitHub.

Step 6: Switch, update and remove layouts

Herdr keeps your existing setup untouched when you add a new layout, so they pile up over time. A few habits keep things tidy:

• Switch layouts with Ctrl+Shift+L. The official guide lists this shortcut for moving between layouts. I use it to cycle through everything installed.
• Remove what you do not use. I had a Windows 98 style theme I no longer wanted, so I asked the agent to uninstall it. It removed it a few minutes later while I worked in another tab.
• Restart to apply updates. Herdr checks for new versions itself. When one is ready, a restart applies it, and your agents keep running in herdr-core in the meantime.
• Review file changes in the app. Herdr can show the files in the current directory and the changes your agent has made, which is a quick sanity check before you commit anything.

Step 7: Use Herdr to build a small Windows app

The best test of a tool like this is real work. In the video I use Herdr with Claude to build an open-source Windows app I am calling Update Forge for now. It scans the software installed on your computer, finds anything with an available update and upgrades it.

Windows already has a command-line route for this with winget, the Windows Package Manager. I wanted a simple graphical interface on top of it. Update Forge uses winget under the hood. I believe the agent built it on .NET, and I did not write any of the code myself.

I show this step on screen in the video, including one detail that tripped me up. When I asked Herdr to launch the app, the agent opened it inside Windows Sandbox, the disposable test environment built into Windows. I had tested the app in the sandbox earlier, which is a sensible precaution for a tool that installs updates. This time I wanted it on my own system, so I told the agent directly: launch the app on my system and do not use the sandbox.

Once it ran on the host, the app scanned my installed applications, found three that needed updates and started upgrading them through winget. That is the kind of small, specific utility a coding agent handles well when you give it a clear brief.

Keep in mind that Windows Sandbox is available on Windows Pro, Enterprise and Education editions. Microsoft does not support it on Windows Home. If you are on Home, ask your agent to test in another way, such as a virtual machine.

If this is the step where you get stuck, the courses and templates inside AI Income Lab cover builds like this one, and you can post your question to the community.

Common mistakes to avoid

• Skipping the AGENTS.md prompt. Without it, your agent may guess at file paths or change the wrong settings. Start each customization by asking it to read the guide.
• Installing mods without looking at them. Mods are code from other people. Open the GitHub link in the gallery and skim what a mod does, and what it reads, before you add it.
• Expecting Herdr TUI agents to appear in the app. The desktop app and Herdr TUI are separate programs. Agents started in one do not show up in the other.
• Assuming agents survive a reboot. Quitting or updating the app is fine. If the machine reboots or herdr-core stops, the agents stop too, and Herdr does not restart them. Start them again yourself.
• Letting the agent choose where to run your app. Say whether you want it launched in a sandbox or on your own system. Otherwise you may end up somewhere you did not intend, as I did.
• Reading usage mods as billing records. These mods read local session data. Your provider’s account page is the authority on limits and charges.

Want a guided path through agent tools instead of piecing it together from videos? See what is inside AI Income Lab.

Is Herdr worth it on Windows?

If you run more than one coding agent at a time, I think it is. Plain PowerShell works for a single session, but it gives you no overview, no status for each agent and no easy way to reshape the workspace. Herdr fixes those gaps, and because customization happens through your agent, you can add the exact panels you need instead of hunting through settings.

It is also a young product with a fast release pace. Expect the occasional broken installer or changed command, and check the official docs when something in this guide no longer matches what you see.

Frequently asked questions

Does Herdr work on Windows?

Yes. The official site lists Windows on x86_64 as supported, alongside macOS on Apple silicon and Linux. You install it with a one-line PowerShell command from the download page, and it adds a Start menu entry.

Is Herdr free?

Yes. The app, herdr-core and Herdr TUI are free and open source under Apache 2.0. The agents you run inside it, such as Claude Code or Codex, need their own subscriptions or API access.

How do I install a Herdr mod?

Find the mod in the Herdr gallery, copy its one-line prompt and paste it into an agent running in Herdr. The agent reads AGENTS.md, fetches the mod’s files from GitHub and adds them next to your setup.

Can I see my Claude Code and Codex usage inside Herdr?

Yes, through community mods. Insights shows token usage by day, week, provider and model, and several mods show plan usage in the status bar or a side panel. They read local data, so confirm limits on your provider’s account page.

Do I need Windows Pro to use Herdr?

No. Herdr itself does not list an edition requirement. Windows Sandbox, which I mention for testing apps, is only supported on Pro, Enterprise and Education editions.

Conclusion

Herdr turns Windows into a comfortable place to run several coding agents at once. Install it with one PowerShell line, point your agent at AGENTS.md, add the mods you actually need, and keep your layouts tidy. Then use it for something real, even a small utility, to see how far a clear brief takes you.