Full Transcript

·YouTLDR

I'm ditching tmux for herdr!

13:59843 summary words · ~4 min readEnglishBy typecraftTranscribed Aug 19, 2026
Analyze another video with Pro30-day money-back guarantee
Summary

Herder is a modern Rust-based terminal multiplexer designed to replace tmux with native, first-class support for managing, monitoring, and delegating tasks to autonomous AI coding agents.

As development shifts toward agentic workflows running multiple CLI AI agents simultaneously, Herder provides built-in agent lifecycle tracking, state-based navigation, and automated pane management without fragile custom scripts.

Section summaries

0:00-0:42

Introduction & Moving on from tmux

optional

The presenter announces he is moving away from tmux after a decade of daily use in favor of Herder. Herder is introduced as a ground-up terminal multiplexer built in Rust that retains familiar multiplexer shortcuts while adding modern innovations.

  • Herder is designed as a direct modern replacement for tmux workflows.
  • It retains standard multiplexing patterns while catering to modern developer tooling.

Gives context for the switch from tmux, but the technical walkthrough starts later.

0:42-2:01

Sponsor: Cursor Cloud Agents

skip

A brief overview and demonstration of Cursor's cloud agents, showing how prepending an ampersand moves command execution to persistent remote cloud instances. The agent spins up development environments, runs local test servers, and returns status captures.

Promotional segment for Cursor cloud agents.

2:01-3:14

Herder Architecture & Installation

watch

The presenter discusses Herder's background, noting it was written in Rust, developed using Zed, and packaged with a Nix flake. He demonstrates the rapid shell script installation method, launches the tool, and shows the default interface with spaces and agent navigation sidebars.

  • Herder can be installed via a quick shell piping script, Homebrew, or Nix flakes.
  • The default interface includes dedicated side panels for workspace and agent management.

Explains the installation methods and baseline interface layout.

3:14-5:52

Keybindings, Configuration & TOML Theming

watch

This section covers Herder's keybinding architecture, which defaults to `Ctrl+B` but can be remapped cleanly in the configuration. The presenter demonstrates the interactive help overlay (`?`), split pane creation, toggling the sidebar in Zen mode, and editing the TOML configuration to apply the Catppuccin theme before using the CLI reload command.

  • Interactive cheatsheet pane is accessible via the leader key plus `?`.
  • Configuration uses clean TOML formatting with instant CLI hot-reloading (`herder server reload-config`).
  • Supports standard pane splitting conventions alongside customizable leader keys.

Crucial for configuring keybindings and understanding daily navigation.

5:52-8:55

Agent Management & State Navigation

watch

The presenter demonstrates Herder's core capability: treating autonomous coding agents like Claude Code and Cursor as first-class entities. He shows how Herder detects active agents, emits audio notifications when an agent blocks on user input, and allows fuzzy navigation with `Leader + G` to filter agents by state (blocked, working, idle, or done).

  • Herder tracks running CLI AI agents automatically across all workspaces.
  • Audio cues notify developers when an agent finishes or gets blocked requiring approval.
  • Fuzzy finder (`Leader + G`) filters agents instantly by state: [B]locked, [W]orking, [I]dle, or [D]one.

Covers Herder's standout differentiator: agent-aware terminal multiplexing.

8:55-11:08

Background Persistence & Agent Skills Integration

watch

This section proves that closing the client leaves background agent processes intact thanks to Herder's persistent daemon architecture. Additionally, the presenter installs Herder's agent skills via NPX, allowing LLMs to directly invoke Herder commands to create new terminal panes and run sub-tasks autonomously.

  • Herder's server daemon keeps long-running agents executing in the background even if the UI is exited.
  • NPX skill installation equips agents with tools to programmatically open panes and execute scripts.

Shows how to integrate autonomous agent tooling directly into Herder's pane system.

11:08-13:13

Remote Sessions with `herder --remote`

watch

The presenter demonstrates connecting to a remote DigitalOcean droplet using `herder --remote user@host`. Herder automatically provisions itself on the target machine and propagates local color schemes, keymaps, and configs over SSH, enabling persistent thin-client development.

  • `herder --remote` automatically handles installation on the remote target via SSH.
  • Local TOML configs and theme settings are mirrored seamlessly onto remote sessions.

Essential for running long-running agentic workloads on cloud infrastructure.

13:13-13:56

Ecosystem, Plugins & Final Verdict

optional

The video wraps up with a review of the growing plugin ecosystem, highlighting add-ons like Herder Navigator, Command Palette, and Smart Tab Rename. The presenter confirms he is fully replacing tmux with Herder for daily development.

  • Community plugins exist for command palettes, navigation, and workspace tab renaming.
  • Herder serves as a full daily-driver replacement for conventional multiplexers.

Quick overview of community plugins and closing remarks.

Key points

  • Native Agent Lifecycle & State Management — Herder includes dedicated UI panes and fuzzy-finders to monitor running AI agents across workspaces, categorizing them by states such as blocked, working, idle, or done.
  • Bidirectional Control via Agent Skills — Herder offers installable skill files via NPX that expose an API for coding agents like Claude Code or OpenCode to directly spawn, split, and run commands inside new Herder panes.
  • Seamless Remote Synchronization — Running Herder with the `--remote` flag connects to a remote server over SSH, auto-installs Herder if missing, and applies the local configuration and color theme to the remote instance.
  • Ergonomic Drop-in tmux Replacement — Built with Rust and configured using standard TOML files, Herder supports classic tmux keybindings, an interactive shortcut overlay, and hot-reloadable server configuration.
Herder is a new terminal multiplexer reimagined from the ground up. Presenter
Herder treats coding agents like Claude, Cursor, tons of other agents as first-class citizens. Presenter

AI-generated from the transcript. May contain errors.

0:00

Hey gang, I have some news to share with

0:01

you. After about 10 years, I'm saying

0:03

goodbye to a close friend of mine. That

0:05

friend, of course,

0:07

is tmux. I'm sorry, was that a little

0:09

too dramatic?

0:10

>> [laughter]

0:11

>> But seriously, check this out. This is

0:13

Herder. Herder is a new terminal

0:15

multiplexer reimagined from the ground

0:17

up. It has all the amazing keybindings

0:19

and features that you would want in a

0:21

terminal multiplexer like tmux, but it

0:23

comes with one little twist. Okay, it's

0:25

actually a really big twist, and I'm

0:27

going to talk about it in this video.

0:28

So, let's get into Herder, gang.

0:31

>> Yes.

0:42

>> Now, before I get into all the

0:43

nitty-gritty details with Herder and the

0:45

setup that I have right here, which is

0:47

awesome, I want to show you something

0:49

really quick that I just learned. You

0:50

see, if you have a cursor agent and you

0:53

prepend a command with ampersand, it

0:55

actually moves that command to a cloud

0:57

agent. You don't know what that is?

0:59

Well,

1:00

let me tell you about this week's

1:01

sponsor, Cursor. Cursor's cloud agents

1:04

are amazing. You can kick off a cloud

1:06

agent like this, and it will start up a

1:08

new development environment in the cloud

1:10

that has all of its libraries, all of

1:12

its context, all of its development

1:14

information needed so I can work

1:16

independently. And the best part is that

1:18

your agent will continue to work without

1:19

your input. I can kick off this agent,

1:21

shut my laptop, go touch some grass.

1:24

I mean, who am I kidding? I'm probably

1:25

not touching grass, but needless to say,

1:28

the agent will keep working without my

1:30

input on its own. And one of the coolest

1:32

parts about Cursor's cloud agents is

1:34

that they return so much information. I

1:37

mean, they don't even just give you the

1:38

results of what they did, you actually

1:40

get back pictures and videos. I mean,

1:42

check this out. This is the initial

1:44

setup to my application, and the Cursor

1:46

agent actually built the app, ran it in

1:49

a local server on its dev environment,

1:51

and started taking a test on

1:53

typecraft.dev. How amazing is this? If

1:56

you want to learn more about Cursor and

1:57

cursor cloud agents, then check the link

1:59

in the description. It's pretty awesome.

2:01

Okay, let's get back to talking about

2:03

Herder. Now, Herder as a piece of

2:04

software is amazing. It is a terminal

2:07

multiplexer as I mentioned before, and

2:09

it was built from the ground up as I

2:11

mentioned before with Rust and I believe

2:15

someone also used Zed to build this

2:17

project, at least one of the developers,

2:19

and it has a Nix flake. Awesome, very

2:21

based. But, Herder is amazing. Now, the

2:24

installation of Herder is actually

2:25

pretty straightforward. There's

2:26

Homebrew, there's a Nix flake, which

2:28

again, very based, and there's a script

2:30

that we can use to install. The script

2:32

is probably the easiest, so let's go

2:33

ahead and use that. We can copy and

2:35

paste the script. It just pipes a

2:37

installation script into your shell, and

2:39

there you go, you're already installed.

2:41

So, Herder is installed and it's ready

2:43

to be run. We can just type Herder and

2:45

there we go. This is Herder. It is a

2:47

terminal multiplexer and it has amazing

2:49

features. Now, the cool thing about

2:51

Herder is that it has basically a

2:52

one-to-one implementation of almost all

2:55

of the Tmux things that you might be

2:57

used to if you're coming to it from

2:58

Tmux. This is something that I really

3:00

love about Herder, but it has a couple

3:02

of goodies that Tmux does not have,

3:05

except for maybe with some plugins. Now,

3:07

you'll see on the left-hand side here,

3:08

there's a little bit of a pane that

3:10

shows my spaces and my agents, but I'm

3:12

going to get to that in a little bit.

3:14

First, let's take a little tour through

3:15

Herder, some of its key bindings, and

3:17

just how it works day-to-day. The first

3:19

thing about Herder you might notice is

3:20

that when you press the leader key,

3:22

which is control B by default, just like

3:24

Tmux, you'll see a little bar pop up at

3:26

the bottom here. This gives you a little

3:28

bit of a preview of some of the key

3:29

bindings you can do and what Herder will

3:31

offer you. My favorite thing is if you

3:34

hit question mark, it opens up a

3:36

floating pane that shows you all of the

3:38

possible key bindings and what they do.

3:41

This is a great way to get started on a

3:43

new piece of software and Herder

3:44

implements this amazingly. There's a lot

3:47

of different ways you can navigate your

3:48

workspaces, your agents, which again,

3:50

we'll get to in a second, and all the

3:52

common uses that you would use for a

3:53

terminal multiplexer like tmux. And so

3:56

you might be asking yourself right now,

3:57

well, if it's just like tmux, can I get

3:59

all my tmux bindings from tmux into

4:01

herder? And the answer is, yes, you can.

4:03

For me, I actually already mapped my

4:05

leader to control S. That is a better,

4:08

more ergonomic way to hit the leader key

4:10

for me. And you can do all the things

4:12

that you would want from tmux in herder

4:14

when it comes to managing panes. And

4:15

another key binding I love is that I

4:17

found you can toggle this pane on the

4:19

left here if you don't want to see it

4:20

all the time. I bound it to leader P. So

4:22

you can toggle that pane on and off, no

4:25

problem. So you have like a little bit

4:26

of a zen mode going on here with herder.

4:28

Right? So we have our new pane over here

4:30

on the right. That's great. That's

4:31

control S percent. And control S and

4:34

double quote will make a new pane on the

4:36

bottom. I know these are kind of

4:37

esoteric and weird commands, but coming

4:40

from tmux, this is just ingrained in my

4:42

muscle memory. And it's nice to see that

4:44

herder can implement this really easily.

4:46

And speaking of configuration files, the

4:48

configuration file for herder is

4:49

actually really easy to read and manage.

4:53

It's built with Rust, so the

4:54

configuration file is a dot toml file.

4:57

If you're familiar with that format,

4:58

it's a pretty easy to read file format

5:00

for configuration. And herder provides a

5:02

lot of different configurations, and

5:04

they're all super straightforward. And

5:06

herder, on top of a great configuration,

5:08

offers a lot of color schemes. You can

5:10

actually choose your favorite color

5:12

scheme by using the theme name space in

5:14

your toml file, and setting the name

5:16

equal to whatever color scheme you might

5:18

want. They implement my favorite color

5:20

scheme, which is catpuccin. So I can

5:22

write this, I can quit, and then I can

5:25

just do herder server reload dash

5:29

config. Herder has a lot of great CLI

5:31

options. And there we go, I've reloaded

5:33

my config. It was already showing the

5:35

catpuccin theme, but if you're switching

5:36

themes, this would switch it for you.

5:38

And again, as I said before, if you hit

5:40

your prefix with a question mark, this

5:42

floating pane will show you all the

5:43

possible key combinations and key binds

5:45

you would use in your normal day-to-day

5:47

workflow. So, it's really easy to

5:49

discover the key bindings that you would

5:51

want to use when it comes to actually

5:52

using Hurder. But, hold on cowboy. If

5:54

you look at the homepage for Hurder, you

5:56

see that they mention coding agents

5:58

right off the gates. This is in the

6:00

first sentence. And that's because

6:02

Hurder treats coding agents like Claude,

6:05

Cursor, tons of other agents as

6:07

first-class citizens. And what does that

6:09

mean? Well, you see Hurder has by

6:11

default this pane on the left-hand side

6:14

of your screen here. You can see it. And

6:15

you can see down below I have an agents

6:18

pane. There's two groups, one for your

6:19

workspaces and one for your agents. Now,

6:22

it's really interesting because Hurder

6:24

does a lot with agents that I think a

6:26

lot of different programs don't do and

6:28

tmux certainly does not do. So, first of

6:30

all, I'm in my dotfiles repo. If I start

6:32

up Claude, I can see here at the bottom

6:34

I actually got a new agent in my

6:36

dotfiles workspace called Claude. You

6:39

see I have another one called Cursor

6:40

right above in my learn directory. And

6:43

there's a lot of cool things that Hurder

6:44

will do with agents that I think help

6:46

with your overall development if you're

6:47

using coding agents. Now, let me just

6:49

show this off really quick. I'm going to

6:50

tell Claude to

6:52

look over my Hurder config

6:56

and make suggestions.

6:59

I didn't type that right, but I'm sure

7:00

Claude will figure it out. Now, I can go

7:02

to a different pane. I can go to my

7:04

browser here and I can just wait.

7:07

Did you hear that? That was Hurder

7:09

telling me that my agent needed

7:11

something. So, I can go back, see that

7:13

my agent needs something. Well, actually

7:15

where is my [music] agent? I navigated

7:17

into a different pane here. I don't know

7:19

where my agent actually is. So, there's

7:21

a couple cool things you can do. Hurder

7:22

is mouse native, so if you are a mouse

7:24

user, which ew, by the way, but if

7:27

you're a mouse user, you can just click

7:28

on the agent and it goes right to that

7:30

agent in your workspace. But, if you

7:32

like using your keyboard, there's

7:34

actually a really, really great key

7:35

binding that I found. Leader G opens up

7:38

a fuzzy finding panel that will go

7:40

through all of your agents and your

7:41

workspaces. [music]

7:42

You can type slash and just start typing

7:44

something like Claude and that will

7:46

fuzzy find the agent that you're looking

7:48

for or the workspace itself. I can also

7:50

type slash.files and it'll go to my dot

7:52

files workspace and I have my pane one

7:54

and Claude right here. But the coolest

7:56

part about this floating window is it

7:58

gives you keybindings to allow you to

8:00

navigate your different agents based on

8:02

the current state. So if I know that one

8:03

of my agents is currently blocked, you

8:05

can see at the bottom right here there's

8:06

some letters and then it says the word

8:08

state afterwards. If my agent is blocked

8:11

and I just heard that sound go off, I

8:12

can type B and it will show me all of my

8:14

blocked agents. I can also type W for

8:17

working agents,

8:18

>> [music]

8:18

>> I for idle agents. My cursor agent here

8:21

is currently idle, D for done agents and

8:24

A for everything. That's amazing. This

8:26

is a super easy way to navigate not only

8:28

your workspaces, but your agents as

8:31

well. And this actually replaces

8:32

something in tmux that I always loved,

8:34

which was the session management. I

8:35

always had it under leader S and that

8:37

would pull up a pane very similar to

8:40

this one that allowed me to fuzzy find

8:41

all of my sessions and to also move up

8:43

and down. By the way, you can use J and

8:45

K in this pane here to go up and down

8:48

all of your panes, workspaces, and

8:50

agents. Navigating everything in Herder

8:52

is super simple with this pane. I love

8:55

it. Also, something really cool about

8:56

Herder and it works the same way in tmux

8:58

is that it runs on a server that runs on

9:00

your machine. So if you close Herder,

9:02

it's actually still running in the

9:03

background. So if you kick off an agent

9:05

and it's doing some work and you close

9:07

Herder, you haven't lost any work. Your

9:09

agent is still going. Let me show this

9:11

off really quick. Now I can go to my

9:13

Claude agent right here. I can just hit

9:15

yes or whatever, yes again, and then if

9:17

I close Herder by typing prefix Q, if I

9:21

look for the process of Claude,

9:25

I can see Claude is actually still

9:27

running right now. That's because Herder

9:29

is still running Claude even though I

9:31

closed the window that Herder was

9:32

running in. Now all I have to do is type

9:33

Herder again, it'll automatically attach

9:35

to where I just was and all of my work

9:37

is still here. The agent was still

9:39

running, it was still doing its thing,

9:41

nothing is lost. This is what I mean

9:43

when I say Herder treats agents as

9:45

first-class citizens. It works amazingly

9:47

with any coding agent you could think

9:49

of. As I mentioned before, Herder

9:50

supports a lot of agents. These are all

9:52

the agents that they currently support,

9:54

and I'm sure they're adding more all the

9:55

time. And along the lines of having

9:57

agents be first-class citizens in

9:59

Herder, Herder actually offers an agent

10:01

skill file. This enables your agents to

10:03

use Herder when it needs to open up new

10:05

panes or workspaces to run tests or

10:07

whatever it may be. It's super easy to

10:09

install. It's a simple NPX command. And

10:12

when you install this with NPX, it

10:14

actually asks which agents you're using,

10:17

so you can install the specific skills

10:20

for those agents. I already installed

10:22

everything I need, so I am going to

10:23

Actually, you know what? I'll install

10:24

OpenClaw. I don't use it, but I'm just

10:27

going to try it. And the installation is

10:28

super straightforward. We proceed, it

10:30

symlinks a skill file to wherever

10:32

OpenClaw looks for these skill files,

10:35

and you will be good to go. Next time

10:37

you use OpenClaw, you can tell it to

10:39

open a pane in Herder and do whatever it

10:41

needs to do. In fact, let me show this

10:43

off for CloudCode. I have CloudCode

10:45

running right now. I can just say, "Make

10:47

a new pane in Herder

10:50

to um write "Hello world." Now, Cursor

10:54

actually loads the Herder skill, and it

10:57

knows how to make a new pane in Herder.

11:00

And there we go. [music] It made a new

11:01

pane. Pretty sweet. If you can't tell,

11:03

I'm still a noob with coding agents. I'm

11:05

trying to learn, but it's been fun.

11:08

Now, let me explain probably my favorite

11:10

feature of Herder. This is available in

11:11

tmux as well, but it's just implemented

11:14

so well. Let me go over a little bit of

11:16

a scenario with you first. Let's say

11:17

you're one of these cool agentic

11:18

programmers. You're working on your

11:20

loops, your graphs, and your agentic

11:22

workflow, but you've got something to

11:23

do. Let's say your wife calls, and one

11:25

of the kids is sick. Now, normally you

11:27

can't leave your computer, your agent

11:29

needs you. If you shut it, that agent

11:31

turns off. But with Herder, you can

11:33

actually put all of your stuff on a

11:35

remote box. You can SSH into it, run

11:37

your Agendic workflow, and then you can

11:39

just leave whenever you want. So, if

11:41

your wife calls and one of the kids is

11:42

sick, you can finally be a responsible

11:44

father again, which is something that I

11:46

look forward to personally. Let me show

11:48

off how this works. Right now, I have a

11:50

droplet on Digital Ocean. I'm going to

11:52

copy this IP address. I'm going to tear

11:55

this down after this video, so don't

11:57

worry about it. And I'm going to quit

11:58

Herder. Now, I can SSH into this remote

12:01

box that has Herder already installed.

12:03

Now, all I have to do is type Herder

12:05

{dash} {dash} remote, my username at

12:09

this IP address. This will SSH into the

12:12

box and run Herder. Now, if Herder is

12:14

not already installed on this machine,

12:16

it will install it in this command right

12:19

here.

12:21

And there we go. It installed Herder,

12:22

and here I am. I'm actually on a remote

12:24

machine. And the coolest part is my

12:26

color schemes and my configuration are

12:29

being used from my local machine. So, I

12:31

can use my great leader key that I love,

12:33

which is control S. The color scheme is

12:35

Cappuccine, which I love. And everything

12:38

works the same [music] exact way that it

12:40

does on my local machine. So, I can use

12:42

Herder on a remote box. I can just have

12:44

my laptop be a thin client, and I can

12:46

use my Herder instances from wherever I

12:49

may be. This is a workflow that I

12:50

personally love, and I'm definitely

12:52

going to set up and start using in the

12:54

near-term future. So, stick around for

12:55

that, because that's going to be a fun

12:57

build. Either way though, now that I'm

12:58

SSHing into this machine, if I kick off

13:00

an agent, I can just get up and leave. I

13:02

can shut my laptop. I can turn it off.

13:04

It doesn't matter, because Herder is

13:06

running somewhere else. All I have to do

13:08

is SSH back into it at another time. And

13:10

finally, I can pick up my kids from

13:12

school again. They've had to walk home a

13:13

lot. Personally, I love Herder. It's

13:16

ergonomic. It makes sense to me, and it

13:18

works with Agendic workflows, which is

13:20

something that I'm trying to get into

13:21

myself. But Herder is huge. This is just

13:24

scratching the the

13:25

They have so many plugins built by the

13:28

community already. Herder is fairly new,

13:30

but look at all of these plugins.

13:32

There's Herder Navigator, Herder Command

13:33

Pallet, Smart Tab Rename. There are so

13:36

many plugins. The community is behind

13:38

Herder. It's a fantastic program, and

13:40

personally, I love it. I'm definitely

13:42

ditching tmux for Herder. As much as it

13:44

pains me to say that, this is just

13:46

implemented so well. I don't see any

13:48

other way. So, check out Herder. It's

13:49

awesome. Subscribe for more Linux, Vim,

13:52

and interesting videos on tech. And hey,

13:54

thanks, as always,

13:56

nerds.

Continue with YouTLDR

Analyze another video with Pro

Process a new video, search every timestamp, compare sources, and keep the result in your library.

Get Pro — $12/month30-day money-back guarantee

More transcripts

Explore other videos transcribed with YouTLDR.