Everything You Need to Know About Grok Bot (Full Grok Build)
574 segments
I'm paying $300 a month just so I could
use Grok bots new viral app. This is
through XAI and now I'm on the super
heavy Grok plan. I'm a big agent user.
I've been using Hermes for a long time.
Before that I was using open claw right
when it came out. My AI agent runs my
entire business. It posts to all of my
social accounts, runs all of my ads and
is my executive assistant. So I wanted
to see how Grok bot stacked up against
Hermes. So I signed up for the $300 per
month super heavy Grok so I could use
this new feature. So today I want to
bring to you exactly what my experience
was with Grok bot and whether it's worth
it for you to set this up and how to
save a bunch of money so you don't end
up wasting money like I did. Now what I
did pretty much as soon as I signed up
for Grok bot is started playing with it
but mostly I wanted to port all of my
existing bot configurations into Grok
bot. So I'll show you how that really
worked as well. And if you stick around,
there's actually a surprise lesson that
I learned through all of this and I
actually ended up changing my entire
Claude code workflow based off of what I
learned from signing up for Grok bot and
it has nothing to do with Grok bot. This
is probably the biggest finding from
this video that you're going to get. So
stick around till the end so I can share
that with you. Now some quick context
just so you know why my answer means
anything. I run several eight-figure
marketing agencies. They're Inc. 5000
companies. I've [music] got thousands of
people inside of my AI automation
insiders community. So I teach this. I
travel the country speaking about AI and
I'm actually using AI to build and run
my [music] business, not just creating
content about it. All right. Now Grok
bot came out August 11th and it's gone
ultra viral. All kinds of creators have
been doing crazy videos about it being a
complete game changer. [music]
And I found some of that to be true,
some of it to be not true. So let's
actually talk about Grok bot and what it
is and then we'll go into actual
implementation and lessons and findings.
So you can sign up for Grok bot on
x.ai/bot
and you'll be able to get some
information about Grok bot. Now the
standout information here, what makes it
really unique, is it's not free. You
need to be on one of either Grok's heavy
paid plans or Cursor's heavy paid plans.
If you'll recall, XAI actually bought
Cursor for like $60 billion pretty
recently and they acquired all of their
coding models, too. So, their coding
models have gotten a lot better. That's
very important for Grok Bot because Grok
Bot runs on these new advanced coding
models. [music] So, here's the site for
Grok Bot. As you can see, this is what
it looks like. Here's my actual Grok
Bot, so you can see it in practice. I'll
be walking through that in just a
minute. It is a app that sits on your
computer. So, I actually have a little
Grok Bot icon down here and when I open
it up, it's just a desktop app. However,
unlike most agents that you've used in
the past where you either need to spin
up a VPS, a virtual server, or run it on
a Mac Mini, with Grok Bot, each one gets
its own dedicated virtual server
immediately. That's why they have to
charge money for it cuz there's server
costs. And here's what they advertise.
It works where you work, so it actually
gets its own browser. It can sign in to
things and take action. You can show
Grok Bot how to do things just like how
you train a Claude skill. You can
actually click around and teach it what
to do. Connecting is really easy and
apparently they get smarter over time. I
have not seen that yet, but I'm sure
it's true. Now, Grok Bot's theory is
that one agent shouldn't do every job.
And this is the theory that I think made
it viral. If you remember Paperclip went
viral, you know, 6 to 12 months ago, all
the agents talking to each other and
coordinating work. Kind of the same
theory. You don't want one agent doing
everything because then it has all of
these crazy prompts and it gets [music]
confused and there's context rot. So,
they advertise you can give each bot a
job and you can have one master agent
that coordinates with all of the other
agents. Now, a couple important points
about Grok Bot. This is my Hermes agent
and I'm texting with it every day. It's
sending me text messages with updates
and status reports. With Grok Bot, you
can only communicate with it through the
app, which actually is really seamless
and easy and fast. So, I actually don't
mind it at all. Whether it's on your
phone or on the computer and it's always
on because it lives in its own server.
Now, the big drawback for most people is
going to be price. So, a lot of you, if
you're using agents, you know, you can
spin up Open Claw or Hermes on a $5 a
per in server or $600 for a Mac Mini one
time and then you never have to pay a
monthly fee again. However, with Grok
Bot, there is a pretty large barrier to
entry that you need to pay in order
>> [music]
>> to use it. Now, they actually just
dropped the pricing. When I signed up,
you could not get Grok Bot for $100 per
month. You needed to be on, I think, at
least their two or $300 per month plan,
but it seems that they've brought that
down since the virality maybe has slowed
down. Now, a couple important things to
note here, you do get access with Cursor
if you want it. However, the big
difference that you're going to see is
this. Significantly higher usage across
chat, imagine, voice, and build. So, you
can either get it through paying Cursor
or paying for Grok. Either one will give
you access to Grok Bot. I recommend
paying through Grok. That way you also
get access to the next thing that I'm
going to share with you a little bit
later in this video, the big
transformational piece for my cloud code
workflow. So, I actually have their
heavy plan, which is $300 per month. I
wanted the full usage of Grok's latest
model. I wanted to fully test it. So,
this is the one that I selected.
However, it looks like now you can get
it through their $100 per month plan.
So, awesome. Once you upgrade and
install Grok Bot, you're going to see
something like this. It looks like a
really simple interface and you're only
going to have this one agent here that
you're talking to. The first thing that
you're probably going to want to do is
connect all of your things. So, down
here you'll see plugins and just like
maybe you've seen my Victor videos in
the past, you can just come through and
connect all of the things that you
normally use in your day-to-day
workflows, all of your marketing tools,
your your operational tools, your Gmail.
And if you want to see the ones that
I've connected, here they are. We've got
Appify, we've got X, we've got Super
base, you've got Higgs Field. So,
they've got a lot of these connectors
already built in. And, of course, you
can bring outside connectors in. So, if
you're like, "Hey, I want to integrate
Hyros." It'll actually collect your key
securely and it'll build an integration
for Hyros for you. So, connections,
really easy, really seamless. Of course,
that's what we expected. However, that's
not a huge advantage over Hermes because
if you are in my programs or you've
watched my videos, you'll know that we
use something called Compozio, which
makes it really easy to connect all of
your apps for free and you can just give
this to your Hermes. So, integrating
plugins and apps isn't such a
differentiator, but it was in the past.
So, they do make it really easy. Then
it's going to walk you through a little
bit of an interview process and Grok
bot, the initial agent that you have set
up, can actually spin up these
sub-agents based on what tasks and
scheduled tasks and tools that you're
using day-to-day. So, as you can see,
it's spun up all of these additional
little bots. I've got my cold email ops,
my long-form desk, my graphic design. It
actually built these itself, built the
descriptions for them, come up with the
the names for them. And if I click into
any one of these, I haven't actually
interacted with this one yet. It's just
a new bot that was created by my master
Grok bot. But each of these, if I click
into it, should have a description and I
should get notifications when it
finishes. So, this is like the system
prompt for what this thing actually
does. Now, in terms of integrations, you
might think that different sub-agents
have their own sets of integrations.
That's not true. They all share the
integrations. So, you don't have to
worry about connecting one integration
to multiple things. Now, what's really
cool is this master agent, if you do
have a master agent like I have here and
I've pinned it. So, as you can see, I
can unpin, I can edit the profile. And
this is really cool. You can actually
see one agent communicating with the
others and you get push notifications
when it's done after it's done talking
to it and getting information from that
other agent. And it's pretty cool to
watch them interacting, see it messaging
back, see a task getting done over here.
But the big question remains, is it
really worth it? Do these sub-agents
actually have a big change in the
results that you're getting compared to
using something like Hermes, which
manages all of my tasks, all of my
scheduled tasks, all of my tools, all
from a single agent. So, I had Grok bot
do a couple of things for me. It built a
presentation, it did some cold email
work, it did some N8N work, it did some
review work, it wrote and posted to some
Reddit threads. All of the things that
my Hermes agent was doing, I had it
temporarily paused while Grok bot took
over. Now, you might be saying, "Jay,
how did you get it trained on all of
these different sub-agents already.
Like, didn't that take a long time? Not
really. And what I recommend that you
guys do once you start using this is
actually have your Claude code. So,
let's go ahead and open up my Claude
code here. I've got my LGJ operating
system. And the first thing that I did
was say, "Hey, I just started using Grok
bot. I would like you to go over my
entire knowledge base, my scheduled
tasks, everything that my Hermes agent
does, and give me a zip folder to train
Grok bot on all of these tasks and set
up sub-agents." And it actually did some
research and it put together a zip
folder. I gave that zip file over to
this master agent and it built
everything for me. Here's kind of how
that initial conversation went. And
also, here's all of the routines that I
had running. And right now, I have all
of them paused. And I'll tell you why in
just a second. I said, "You're going to
be my master agent, coordinate with all
of the sub-agents. You are the one I
will talk to directly and delegate
tasks. I'll be sharing a zip folder that
has all the information and skills and
integrations and tooling for what we
will need. You're going to be taking
over for my Hermes bot." I just told it
I configured a bunch of plugins. How do
I sign into them? And then it prompted
me one at a time to actually initiate
the the login cuz some of them use OAuth
and some of them use API. So, we just
spent some time configuring plugins.
Before it was added, it either prompted
me to sign in like this, authorize, or
asked me for an API key. So, we just
went through, configured everything. I
removed one of the plugins. As you can
see when it needs to do a sign in or I
need to do something in the computer,
this is its screen. And I can actually
click into a screen and see what it's
doing at any given time. This is its
computer that it always has access to.
I'm going to go ahead and minimize that
now so we can keep going. And here is
that zip folder that I dropped to it.
Unpacking now, it's a 12-bot provision
pack, not an import. So, it went ahead
and built all of the sub-agents based
off of this zip folder that I gave it.
From there, we configured all of those
little sub-agents and specialized bots.
And it did it. And we turned on all of
the scheduled tasks and Grok bot started
running. So, all in all, in theory,
sounds really cool. But what I noticed
was it wasn't quite doing as good of a
job as my Hermes agent was. Now, maybe
that is because I didn't spend nearly as
much time training it and refining it as
I did with my Hermes agent. But, from
what I can tell, Hermes is still the
better agent harness. Now, yes, you're
going to need to either deploy it on a
virtual server or get a Mac mini like I
have so I can iMessage it. So, there is
that kind of additional steps, kind of
additional cost whereas Grok bot you can
kind of just launch it, it already has
access to servers, you don't have to
worry about it. You don't have to worry
about integrating an LLM cuz it just
works on Grok. However, I've still found
my Hermes model to be better than Grok
bot. So, this is Bob, I text it, this is
my Hermes agent. He lives on my Mac
mini. We also can deploy these really
easily using a VPS. In fact, if you're
in my AI Automation Insiders community,
we literally deploy these for you in one
click. So, this does everything. It's
connected to all of my tools through
Compozio. It sends comments and DMs
across all of the social channels,
monitors all of my ads. If I tell it to
do anything, it's able to pretty much
power through and figure it out. It uses
my Codex subscription, which I have a
$200 per month subscription, but you can
run Hermes on a $20 per month chat GPT
subscription. And that's pretty much it.
Or you can get really frugal and get
like open router or an open source model
and run it on something cheaper. But, it
does a great job running on Codex. I
also really prefer the way that Hermes
handles context windows, auto
compaction, scheduled tasks. It just
does a really nice job and manages
everything really well. But, the
question remains, is that one agent
responsible for all of those tasks done
it going to do as good of a job at all
of those different tasks as Grok bot
will since it had it can delegate all of
those tasks to sub agents. Now, I am
somebody who does a lot of tasks.
There's a lot of different scheduled
tasks, there's a lot of tools, there's a
lot of different things that I ask it to
do. So, I feel like I would be a pretty
good use case of determining if it was
capable of it or not. And I'm here to
tell you, yes, Hermes agent is just as
capable as Grok bot at executing
multiple tasks, doing multiple things,
all without having to delegate to a
thousand different sub agents. And most
of you aren't doing things that are that
complex. You don't have a million
different tools or businesses that
you're overlooking. You're just using it
for pretty simple use cases, social
media automation, executive assistant
stuff. This is my Hermes agent running
on my Mac mini. This is the user
interface. If you ever want to log in
and see it, they do have one. And Hermes
actually does have the ability to
delegate tasks to subagents through
[music] Kanban and through agent
orchestration. They have this built in.
So, it's essentially the same thing that
Grokbot is doing just behind the scenes.
All right. Now, before I go into what
actually really works and what I
actually implemented, and it's not
Grokbox, I'm probably not going to be
using that long term. I'm going to stick
to Hermes. It's a great solution. And I
still think it's the best possible
solution for most of you that want to
train a bot over time. Now, if you don't
want to worry about setting up a VPS or
getting a Mac mini, we can actually
deploy them for you and just fill out
this form and within a minute you'll
have a Hermes agent that you can message
back and forth that just works for you.
It also comes preloaded with all agent
skills. Or if you don't want us to host
your Hermes agent for you, that's fine,
too. You can go to legionai.com/bob.
Bob is the name of my Hermes agent, it's
what I call it, as you can see from my
text messages. I'm messaging back and
forth with Bob. This way you can just
copy our configuration into your own
Hermes agent so that you can host it
yourself on whatever platform that you
want. And for those complete
do-it-yourselfers, just know Hermes is
free. It's at this GitHub repo, Hermes
agent. It's updated all of the time.
It's got 235,000 stars. It is probably
the most popular and most powerful AI
agent harness at this time. And in my
humble opinion, after trying multiple
different agent harnesses, open claw,
paper claw, Hermes by far my favorite,
by far the best. However, [music] if you
do plan on subscribing to one of Grok's
heavier plans, which you might, then you
might not need Hermes agent. And now I
might just talk you into subscribing to
one of Grok's paid [music] plans,
anyways. Now, here's why I recommend
subscribing to Grok, whether or not you
decide to use Grokbot. If you already
have a Hermes agent and it's working
great, awesome. If not, you might want
to sign up for Grokbot anyways because
of this. Grok's latest model 4.6 just
came out and it's significantly better
than GPT Soul in most use cases. If you
just click through most of the
benchmarks, you'll see that it's pretty
much just crushing them and I think
they're going to continue to crush them.
They're also extremely generous with
credits just like ChatGPT is. Why is
this important? Since you're probably
using Claude code, I'm using Claude code
as my primary agent driver. So, let's go
ahead and open up my Claude code. So,
now that we're in my terminal, this is
running on tmux, but what's important is
this is all running Claude code. All of
these sessions are Claude code. These
are actual real sessions that I'm
running right now. Now, before signing
up for Grok, I had these two skills that
I would use pretty frequently. They were
{slash} codex review where I would
actually have GPT 5.6 Soul where it
would actually go back and forth [music]
with Claude to do an adversarial review
and make sure that the plan that we're
coming up with and reviewing is actually
good. And it was a game-changer. It
absolutely caught way more bugs than
Claude did on its own and really
improved my developing process. However,
GPT 5.6 took a long time to go back and
forth and review with Claude. It would
miss things all the time and then it
would catch things and it just wasn't
the best possible co-pilot. I thought it
was at the time and then I would
actually use another command called
codex execute to actually write the
code. Since it's much cheaper than using
Fable to write code, it would use GPT
5.6 Soul to actually write the code and
make the fewest mistakes. And this ran
on my $200 per month ChatGPT
subscription. So, it saved me a ton of
money in Claude code and stopped me from
hitting my limits. And by the way, guys,
if you want access to any of my Claude
code skills, you can go to
leadgenjay.com/skills
and I put all of my most important ones
here so you can come in here at any time
and kind of see what I'm actually using
in my day-to-day and install them for
yourself. So, if any of these commands
look interesting, they're all in here.
Anyways, now that I had this $300 per
month super generous Grok subscription
with all of these new Grok credits. I
needed to see how well this could
perform, how good it was at coding. So,
what I did is I ran a head-to-head and I
just posted about this in my AI
automation insiders community. It is
free if anyone wants to join. This is
where I give the latest updates, all of
my tactics, all of my skills, you can
download them all within here. But, I
posted this a while ago and actually
showed the benchmarks. Here we go. So,
let's actually review these now. Grok
did the same task in 12 minutes 53
seconds compared to about 30 minutes
with Codex. And that's with 45 test
passes compared to Codex's 33 test
passes. So, it was more than two times
faster and more accurate and ran more
tests. Now, in order to use this inside
of your Claude code, you need to install
the bridge. And this bridge is actually
a free GitHub repo that you can just
install. It's from Grok. So, it's this
Grok build plugin CC. And this is the
Grok build Claude code bridge. So,
you'll give Claude this GitHub repo and
it'll install all of those skills and
commands that I use inside of my actual
Claude code terminal. So, now when I'm
doing anything new, I'm actually going
to type {slash} Grok and it will call
Grok on my paid subscription to review
and write the code. So, now instead of
just using Fable for everything, cuz
yeah, if you look at the benchmarks,
Fable still crushes everything, but it's
a token consumer. So, now I can use Opus
5 and have it review and write code with
Grok's latest model, which is
significantly better than Opus. And I'll
get a much better and more accurate
result. So, that was my big takeaway
from this whole Grok experiment. Now,
I'm probably not going to continue to
use Grok bot. However, I probably am
going to keep my Grok subscription. I
don't know whether I keep the $300 or
downgrade to the $200, but it is proving
so far to be worth it. Mostly as a
Claude code copilot. So, I hope this
video was helpful. Hopefully, it saves
some of you some time, some money, some
experimentation. And if you want us to
deploy a Hermes agent for you,
pre-trained, ready to go, you can join
our AI automation insiders community.
The link is going to be down below. It's
also got live coaching with me, tons of
lessons, tons of modules. We go really
into the weeds about specific topics.
Hope to see some of you in there. If
not, I've got a free community that you
can join as well. I'll also link all of
the resources that I talked about in
this video down below in the
description. And finally, I'd appreciate
if you hit the like button. Show your
support. Thank me for making this free.
And in the comments, I'd love to hear
any of your feedback. Have you
experimented with Grok bot? What do you
think about it? Make sure to subscribe
to the channel if you love AI for more
like this. I'll see you soon.
Ask follow-up questions or revisit key timestamps.
The video details the creator's experience with the newly released Grok bot, comparing its performance as an AI agent against his existing 'Hermes' setup. While Grok bot offers a unique, dedicated server architecture and the ability to spin up specialized sub-agents, the creator concludes that Hermes remains more capable and flexible for his professional workflow. The most significant finding, however, is not the bot itself but the integration of Grok's latest coding model into his existing 'Claude code' workflow, which resulted in faster and more accurate coding tasks.
Videos recently processed by our community