So, let's be honest about something. In

its current state, Hermes Agent is a

better, more reliable AI agent than Open

Claw. They've added a ton of

functionality over the last month that

has not only made it a super powerful

24/7 AI employee, but more importantly,

has made it incredibly reliable. In this

video, I'll go over all seven new

changes that I absolutely love, show you

how to use them, and make you a master

of Hermes Agent. Let's lock in and get

into it. So, this is the brand new

Hermes dashboard I'll be walking you

through. I'll be taking you through a

bunch of really incredible features I've

been using that I love, including what

you see here, which is the brand new

Kanban board. I'll show you some amazing

workflows so you can be doing a ton of

things at the same time. Here's the deal

with Open Claw real quick. Let's address

the elephant in the room. Open Claw has

had two major issues over the last month

that has led to a tremendous amount of

people switching from Open Claw to

Hermes. You check Google Trends anything

right now, the chart for Open Claw is

going like this, the chart for Hermes

Agent's going like that. The two major

issues are that one is, and this is the

part that pisses me off the most, every

single update breaks Open Claw. They

release an update every day. Every

They're shipping a ton. I'm not going to

hate on them for that. They're shipping

a ton, but everything they ship breaks

Open Claw. Every time I update Open

Claw, it breaks and I have to go back in

and spend half an hour fixing it. That

is not an experience people want. The

second thing, and I'm seeing this

probably more often on the timeline, is

bloat. The app is just getting huge,

it's getting too many features and

functionality, and the more people use

it, the more it slows down over time.

Performance has just crumbled a ton for

people. For me, I experienced this once

a few months ago, then I spent hours and

hours and hours trying to figure out why

it was happening and what do to prevent

it from happening again. I found the

culprit to mostly be the way it manages

sessions. I have another video from a

couple months ago when I solved this on

managing your sessions. So, performance

for me in Open Claw hasn't been as much

of an issue, but for a lot of people it

is, but the broken updates have been an

issue, and because of that, I can't

really enjoy any of their new features

cuz I'd rather stay on an old update

than going on a new update and have to

spend a half hour every single day

trying to fix it. So, with that being

said, we are using Hermes Agent now. I'm

going to go through it all here. The

Hermes Agent team seems to be very, very

focused. They are doing fewer updates,

but the updates are much more reliable

and seem to be all fitting into specific

narratives. So, it's not just 100 things

crammed into one update, it's several

very specific updates that make sense.

You can see here every update, which

again, by the way, their aesthetics,

their marketing team, incredible. Props

to them. All have very specific themes,

they're all focused. So, when you

update, you know what you're getting.

You're not getting just 100 things

thrown at you and they all break, you're

getting very specific updates that fit

into a narrative. So, you understand

what's happening with your agent, which

is very important. Shout out to the

Hermes team just for being focused like

this, it's great. Anyway, let's get into

it. Let's talk about Hermes. So, this is

the Hermes dashboard. I never

helpful, but there are some amazing

things in this Hermes dashboard. All you

need to do to get this running is you

open up your CLI and you type in Hermes

dashboard. We're in here. Hermes

dashboard, you hit enter, it gives you a

URL where your dashboard lives. Now,

once you're in here, there are many

things you can be doing. I'm going to

start off with one of the biggest

features they released over the last

month, which is the Kanban board. This

is absolutely glorious for multitasking

with Hermes Agent. When you are talking

to your Hermes through Telegram or

iMessage or Discord or whatever it is,

you are talking in a single thread. You

send a message, it receives that message

and does work. If you try to send 10

messages at one time, they all get

queued. So, it's a very single file way

to do work. On the alternative is the

Kanban board. Now, you can do 10, 20, 30

strands at the same time. And if you're

anything like me and you work on 20

different things a day, you're building

a SaaS, you're writing content, you're

managing community, you're doing a whole

bunch of different things, you can't

just have a single thread. You need to

have multiple things going on any given

time. This is where the Kanban board

comes in. And this is great because it's

deeply integrated with everything

Hermes. Let me show you how this works.

So, what I like to do is when I come in

at the beginning of the day, I go into

triage over here. So, that is the

leftmost column. There's a bunch of

columns here, triage, to do, ready, in

progress, blocked, done. It's all in

order, it's all how the tasks flow

through. We start off in triage. I go

in, I have my to do list for the day,

you can see it right here, and I go in

and I type in everything I need to do.

So, I'll say discuss him hiring with AI,

create, I'll do another one, message

live stream clippers, create that, and

then plan second channel. By the way,

link for the second channel down below,

Alex Fin Labs, check that out. Create

that. And so, this puts all my tasks in

here. Now, what's going to happen is in

the next 60 seconds a cron job will

fire, and I have a Hermes Agent. I'll

show you how to do all this in a second.

I'll show you how to set up everything

I'm doing here. I have a very custom

workflow I'm showing you right here.

Within 60 seconds, my agent that's

dedicated to managing this Kanban board

will go in, see what's in triage, look

at all my memories, fill out these tasks

with details based on my memories so

it's relevant for what I want to

accomplish, and move it into ready.

Allow me to then take that fully fleshed

out task and assign it to an agent so I

can get that work done. So, you look in

here, I created all these tasks a few

minutes ago, they've all been taken from

my manager agent, fleshed out, and put

into ready. So, let's pop open one of

these here. You can see here the agent

went in, looked at all my memories from

my Obsidian memory system. I had a video

on that like a month ago if you want

information on that, and took all the

context around it and put all these

details in here. So, instead of me

having to go in and sit here for an hour

typing all the details of what I want

the agent to do and add all of this

context, my agent is automatically doing

that for me. Now, all I need to do is go

in here, assign it to one of my agents,

my main worker agent just named Hermes.

I just like the name Hermes, I didn't

give it a creative name. It is what it

is. I hit enter. Now, this is assigned

to my main Hermes agent who will then in

a second do the work for me. Look at the

task name, look at the information, take

it from ready to in progress, and manage

it all the way to done for me. This is

the best way I've seen multitasking with

agents ever, and it is native and

directly inside Hermes Agent. All that's

required out of me is two things,

creating the tasks in triage, and then

assigning it when they're in ready.

That's it. All the work is done for me

by my agent team. You can see here,

boom, I as it was live doing this video,

that task was automatically picked up by

Hermes Agent and it's gone and moved

over to in progress. And what's amazing

is as the agents do different things, as

they fill out the details, do the work,

they add comments down here. So, you can

see the librarian, who's my manager

agent, spec written by the librarian,

output will land in the Obsidian Vault.

I have an entire folder in the Obsidian

Vault, which I'll show you in a second,

that all this output goes into. So, how

do you set this simple. By default,

there's no automation here. By default,

there's no agents picking things up,

adding details. I created this workflow

completely custom by myself. I'll show

you how to do that in a second, but by

default, you're just putting in tasks

and assigning them, and eventually they

get picked up by the agent. Here's what

you want to do to enable all of this

amazing automation that will save you

tons of time. Here's the prompt, and

I'll put this down below. I want to work

on the Kanban board inside the Hermes

dashboard. Please create a cron job that

checks the Kanban board every 10

minutes. I decided 10 minutes is the

sweet spot. You can go like every 30

seconds if you want. If you create a

agent that manages this, that's on a

super cheap model. I mean, even if

you're on an expensive model, it's not

really a big deal. It's not using a ton

of tokens. So, you can do it quicker if

you want. 10 minutes is fine for me. To

see if there any task in triage, then

take the tasks, look through all the

memories, and fill out the details for

the task, then move it into ready. Then

check to see if there any task in ready

that I assigned an agent to. If I did,

take care of that task. This prompt, if

you take it from down below, put it into

your agent right now, will get you all

set up. You have two different options

here. You can have your main agent, your

Hermes agent, managing this. I'd

actually recommend against this. I would

recommend setting up a second Hermes

agent that will manage this. That's why

I have a librarian. This is my

librarian. It's running on Chat GPT, my

OAuth, so I'm not spending any extra

money on it, and I have it do a lot of

my administrative tasks. If you just go

to your main Hermes agent and say, "Set

up another Hermes agent profile for a

librarian," or whatever you want to name

your kind of administrative assistant,

it will set that up for you, and then

you just tell your main agent, "Hey,

have the librarian check every 10

minutes." The reason why you want to do

that is if your main agent's doing

something every 10 minutes, checking the

board, having that cron job, it can

block tasks, right? It could slow down

your workflow. So, you want a second

agent that's doing that every 10-minute

work, and that's what the librarian

does. If you set this up and you have

this going all day and you put your

tasks in and you manage it, you're going

to automate your workflows and you'll

get more done you ever have before. The

next major feature I want to talk about,

and I think this is one of the most

underrated developments in AI this

entire year, is {slash} goal. There are

two major platforms right now that

utilize this new {slash} goal

functionality. That's Codex from OpenAI

and now Hermes. Hermes actually beat

Open Claude to this. Let's talk about

{slash} goal and how it works. So, let's

go into our terminal. We're going to say

Hermes. That is going to open up our

Hermes CLI so we can interact directly

with our Hermes agent. From in here, I

can now use {slash} goal. {Slash} goal

is very different than just sending a

prompt. When you send a prompt, it has

very specific actions and the agent

takes those actions. {Slash} goal

instead gives your agent a mission. It

is a high-level mission that the agent

can take many steps to do that can go

over very long time periods. I've seen

some people do {slash} goals that have

literally run for three plus days

straight. Obviously, you don't want to

be using Opus for this if you're doing a

multi-agent mission. I'd probably

recommend using the ChatGPT OAuth or

even better, a local model. Video on

local models coming very soon. But,

{slash} goal allows you to do very

high-level big missions that your agent

will go after, figure out how to

accomplish, do multi-step approaches,

test itself the entire way. It's

basically like a Ralph loop but for your

AI agent. So, you got to open up the

CLI, you got to do {slash} goal, and

then you type in your goal. One thing I

need to say here is the quality of your

prompt for the {slash} goal is very,

very critical. If you just do say

{slash} goal and say, "Hey, build me a

game." you're not going to get very good

results. I would highly recommend meta

prompting here. And basically what I

mean by that is go to your favorite LLM,

ChatGPT, Claude, whatever, say, "Hey,

I'm trying to build a {slash} goal

prompt for my agent. This is what I'm

trying to accomplish. Please write me a

prompt that gives me a super detailed

{slash} goal prompt." It will give you a

prompt that will be probably extremely

detailed and allow your agent to do a

lot more work. I did that already. I

have a prompt right here where it's

going to audit my entire Obsidian vault,

all the memories I have inside it, find

the dead memories we have and update in

a while, see which ones are kind of

orphans that are not in folders, and

give me a cleanup plan that I can then

approve and it goes and cleans it up.

So, I'm going to hit enter on this and

my Hermes agent will go and build an

entire plan and then clean up my

Obsidian vault. It'll go on this mission

and it can run for minutes, hours, days.

It will just figure out the best way to

accomplish this high-level mission. I

highly recommend doing this. If you

don't off on the top of your head know a

goal you want to give your agent, I'd

recommend doing this. Go to your agent,

say, "Hey, I want to give you a {slash}

goal, but I don't know what goal I want

to give you. Take a look at our

projects, things we've done lately, and

come up with three options for goals we

can execute on." Then you choose one of

those goals, and then it goes and it

tackles that goal. So, I'd highly

recommend doing a little bit of reverse

prompting here if you're not sure what

you want to do with {slash} goals, but

it is a great feature that goes and

sends off your agent and does amazing

work. If you learned anything at all so

far, we have like four more features I

want to go through. This is going to be

a wildly helpful video. Learn anything

so far? Leave a like down below. Make

sure to subscribe, turn on

notifications. Also, join the Vibe

Coding Academy. I have full video series

on Hermes and Open Claude and a ton of

people in the community who are experts

on these tools. You'll learn a ton. I

promise you it'll be the best thing you

ever do. Join the Vibe Coding Academy,

link down below. Next, I want to talk

about profiles, aka multi-agents. Hermes

put a focus on making it so you can

create multiple agents very easily,

create teams of these agents that you

can then customize any way you want and

start attacking tasks on your Kanban

board. Much easier to do than Open

Claude. The way you do this is you can

either ask your agent create new agents

or you can go in the dashboard here,

click profiles, multi-agents, go in

there, give it a name, name your new

profile, and click create and boom, you

have a new Hermes agent. This is its own

Hermes agent with its own memories, its

own skills, its own environment

variables, everything unique to it. The

reason why you want to have multiple

Hermes agents is if you have one agent

doing many, many things, coding,

writing, research, its memory will get

very bloated very quick and it will hurt

its performance. But, if you make

multiple agents, they can have their own

individual memories and skill sets so

things don't get bloated and performance

stays high. So, you want to go and you

want to set up agents. I'd highly

recommend having a coding agent. I'd

highly recommend having a research

agent. And then I'd highly recommend

having a administrative agent like the

librarian which we were going over

earlier. And then also a main agent.

Mine's named Hermes. This is kind of

your general orchestrator that you'll be

messaging on Telegram or whatever that

will be deciding which Hermes agents to

send on different tasks. I think those

are kind of bare-bones basics if you're

anything like me and you do a lot of

different things on a daily basis. But,

you want to be setting up your agents in

this new profile section. It makes it so

easy to set up new agents. Then you can

go in, go in your Kanban board, start

assigning them tasks, which is really

cool. Next, let's talk about the model

catalog. They in here have models. And

what I love about this is it simplifies

a action we've been doing in agents for

a long time, switching models. It makes

it a lot easier for people to do, track

your costs, things like that. So, you

can come in here and you can quickly

switch out which models you're using.

And what's amazing is you can even go in

and you can select specific tasks to

assign that model to. So, if you've seen

my past videos, I talk about the brain

muscle model, right? Having an

orchestrator model and then muscles with

cheap models on each. You can easily set

that up here. So, if I want to say GPT,

which is a little bit cheaper than Opus,

I want to use that for any sort of

approval task, MCP, things like that, I

can do that right here, which is great.

It makes it really easy to set up the

brain and muscle model. So, model

catalog really, really nice and it helps

you track all your costs, too, which is

really cool. One other thing I want to

talk about, too, here, a new feature is

compression. This might not actually be

a new feature, but it's a feature I'm

taking advantage of now. The one

complaint I have about Hermes that kind

of bothers me is it seems like their

memory compressions are very violent.

And what I mean by that is whenever

there's a compression, it seems like it

totally forgets everything, which is

very annoying. One thing that's improved

that a lot is by lowering the

compression threshold. Basically, what

it means is it does more compressions,

but they're not as dramatic. There's not

as many things to compress. So, if you

go to config, compression, and then

change your threshold to 0.5, you'll do

compressions

they won't be as dramatic or violent.

They'll be compressing smaller amounts

of data. This for me has helped improve

the compressions a little bit. I do

believe this is still something that is

weaker on Hermes than Open Claude is the

kind of memory management and

compression, but this has helped a

little bit. Recommend config,

compression, change it to 0.5 for the

threshold. One other feature I'll talk

about they added kind of quietly in the

background just recently is the curator

feature. This is a great one. This is

actually not something you have to

actively use. It's happening

automatically. I just want you to be

aware of it. What happens is is this is

actually very helpful. This is something

probably Open Claude wants to put in to

help remove bloat. This is an action

that happens every 7 days by default.

And what it does is it looks through all

the skills you're using. It makes report

on the most used skills and the least

used skills. And it will actually prune

out skills you're not using. So, it

reduces the bloat of Hermes. It even

builds a report that you can check out

to see which skills you're using, which

ones you aren't. So, if you want to

manually change things, you can do that

as well. This is how Hermes maintains

itself and self-improves itself. It's

just always pruning and curating the

skills and the memories it's using. This

prevents that kind of bloat that's

really slowed down Open Claude over the

last month or so. Really cool feature.

You can go into your config, curator,

change anything you want here. If you

want to change it from going every night

instead of every 7 days, you can do

that. I think every 7 days is a pretty

good sweet spot, but the curator feature

is fantastic as well. These kind of

targeted, strategic features they're

adding, I think have made Hermes just a

lot more reliable of an experience.

While it seems on Open Claude they're

just kind of implementing a hundred

different things every single day, which

is why things are constantly breaking. I

mean, I get it. Peter Steinberger was

hired by OpenAI. He's probably

distracted by a bunch of different

things. I've seen the timeline. There

are a ton of complaints. I'm hoping they

improve things over time and hear the

complaints. I I'm confident they are. If

you implement everything I showed you

here, you're going to get so much more

out of your AI agent. There's a lot more

coming soon on Hermes. Let me know down

the replies what you'd like to see next.

Do you want to see more AI agent videos

like Hermes and Open Claude? Do you want

to see multi-agent approaches with the

two? Do you want to see more Vibe Coding

stuff with Codex and Claude Code? Do you

want to see more agent stuff around

Claude Co-work? Let me know in the

replies. Join the Vibe Coding Academy as

well. Tons of video series there on

Codex, Claude Code, Hermes, Open Claude,

all that. Huge 1,500 person community

there. You'll love it. It'll be a great

decision. Also, starting a second

channel. It's called Alex Finn Labs.

There's going to be a ton of really

amazing content there. I'm very

confident you'll love. Link for that

down below. Thank you for watching the

video. I'm so deeply appreciative you

take time out of your day and spend it

with me. I will see you in the next

video.