Big Idea

The argument in one line.

A Grok Bot agent is defined by its job description, a shared cloud computer, and its own memory, and those three pieces are what let dozens of bots work as an org chart instead of a flat list of assistants.

Who This Is For

Read if. Skip if.

READ IF YOU ARE…

Someone running multiple AI agents or chatbots who wants them to delegate work to each other instead of operating as isolated assistants.
A solo operator or small team lead who wants one shared AI workforce with its own computer, memory, and task history.
Anyone evaluating Grok Bot against Claude Code, OpenAI's agent tools, or a custom multi-agent setup and wants the concept list before signing up.

SKIP IF…

You just want a single chatbot to answer questions; this is about running a team of autonomous agents, not one assistant.
You need technical build detail on how to construct a multi-agent system yourself; this is a tour of an existing product's UI, not a build tutorial.

TL;DR

The full version, fast.

Grok Bot is xAI's multi-agent platform: every bot you create gets a name, a job description, and its own memory, and all your bots share one cloud computer with a browser, file explorer, and terminal, plus access to your local files. Plugins connect accounts like ClickUp and Gmail, skills are reusable slash-command recipes you can teach by recording a screen demo, and routines fire those skills on a schedule or off a webhook. Bots delegate to each other by mention inside group chats, pause for human takeover on sensitive logins, and log their own work to an external tracker so a growing team of agents never becomes a black box.

Free for members

Chat with this breakdown — free.

Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.

Create a free account →

Chapters

Where the time goes.

00:00 – 02:02

01 · Grok Bot Basics

Each Grok bot is a single agent with a name, job title, and description, and that description is what lets other bots know when to delegate work to it.

02:02 – 05:20

02 · Templates, Context & Memory

Bots can be exported and imported as shareable templates, each one keeps its own conversation context plus a private memory store on top of a memory every bot can read, and long conversations auto-compact instead of hitting a context limit.

05:20 – 08:05

03 · The Agent Computer

Every bot works on one shared cloud computer with its own browser that stays logged into sites, a file explorer, and a terminal, and bots can also read, rename, and move files on your actual local desktop.

08:05 – 12:08

04 · Plugins, Skills & Routines

Plugins connect bots to real accounts like ClickUp, Gmail, and GitHub with support for multiple accounts per plugin, skills are reusable editable recipes you can also teach by recording yourself doing the task once, and routines run a skill on a schedule or trigger.

12:08 – 13:46

05 · Webhooks & Event Triggers

A webhook wakes up a specific bot when an external event fires, like a form submission, so it can process the payload and take action across email and a task tracker without you starting the conversation.

13:46 – 16:24

06 · How Bots Work Together

Bots can sit in shared group chats, mention each other to loop someone in, hand off work directly to another bot, and a reply button isolates one sub-thread so follow-up instructions don't get misapplied to the wrong message.

16:24 – 19:21

07 · Controls, Usage & Monitoring

Approvals gate what a bot can do on your local computer, bots pause for human takeover on sensitive steps like logins, weekly usage resets every seven days with an on-demand spend option, and run history shows when routines fired.

19:21 – 20:35

08 · Managing Your Bots

Per-bot notifications sync to your phone, and bots can be pinned, moved into folders, hidden, duplicated to reset conversation history while keeping skills, shared as templates, or deleted.

20:35 – 21:22

09 · Agent Logging

The bonus concept: have every bot log its own actions, progress, completion, archive, to an external system like ClickUp, a spreadsheet, or a database, so visibility doesn't disappear as the number of agents grows.

21:22 – 21:32

10 · Free Grok Bot Course

Closing pitch for a free two-hour Grok Bot course.

Atomic Insights

Lines worth screenshotting.

A Grok bot's job description isn't documentation, it's the routing logic other bots use to decide who to delegate work to.
Every Grok bot shares the same cloud computer, so a file one agent creates is instantly visible to every other agent's file explorer.
Grok Bot keeps two separate memory stores per bot: a global memory every bot can read and a private memory only that bot keeps.
Conversations auto-compact with no user control, so a bot talked to for months will summarize old messages instead of forgetting them outright.
Recording yourself completing a task in the browser lets Grok Bot turn that screen capture directly into a reusable skill.
A routine is just a skill plus a trigger, either a time schedule or an event like a webhook, Slack message, or form submission.
Bots pause for human takeover on sensitive steps like logins, because anything typed into the conversation history isn't protected the way a password field is.
Duplicating a bot resets its conversation history while keeping its skills and routines intact, which is the fix for an agent that's drifted off track.
Weekly usage resets every seven days, and a multi-agent group chat burns through that quota faster than a single one-on-one conversation.
Sharing a bot as a template hands over its instructions, memories, and skills, but not its connected accounts, those still need re-authenticating.
Agent logging turns every bot action into a row in an external tracker like ClickUp or a spreadsheet, the only way visibility survives scaling past a handful of bots.

Takeaway

The Building Blocks Of A Multi-Agent Team

AGENT TEAM BASICS

Running more than one AI agent well comes down to clear job descriptions, a shared memory and computer, reusable skills, and logging every action so visibility doesn't disappear as the team grows.

01Grok Bot Basics

A bot's description is read by other bots to decide who should handle a task, so writing one specific, honest job description matters more than naming it well.
Scaling past a few bots works better as a reporting hierarchy, a few 'executive' bots who know their team's strengths, than as one flat list you manage yourself.

02Templates, Context & Memory

A shared bot template carries its instructions, memories, and skills but not its connected accounts, so the recipient still has to reconnect their own tools.
Treat global memory and per-bot memory as two different scopes: update global memory for facts every agent should know, and a specific bot's memory for context only that agent needs.
Long-running agents auto-summarize old messages instead of forgetting them, so don't assume an agent you've talked to for months has lost the thread, it just isn't showing you the full history.

03The Agent Computer

Because every agent shares one file system, dropping a file for one bot to use makes it instantly available to every other bot, no manual handoff needed.
An agent that can log into a site and stay logged in can act inside your actual accounts once credentials are saved, so treat that access with the same care as handing someone your password.
Local file access means an agent can rename and reorganize files on your real desktop, not just inside its own sandbox, worth confirming before trusting it with anything irreplaceable.

04Plugins, Skills & Routines

A good skill documents when to use it, the exact order of steps, and what to do if something goes wrong, the same structure as a recipe you'd hand to someone else.
Recording yourself completing a task once and letting the agent turn it into a skill is faster than writing instructions from scratch, especially for browser-based workflows.
A routine is only a skill plus a trigger, so the real design decision is picking the right trigger, a time schedule for recurring work, an event for reactive work.

05Webhooks & Event Triggers

Webhooks let an agent react to something happening outside the conversation entirely, a new form, a new email, a new CRM record, instead of only responding when you type to it.
A single webhook event can fan out into multiple actions, an email reply, a task log, a teammate notification, if you tell the agent what full response you expect.

06How Bots Work Together

Group chats let multiple specialized agents reason about one open-ended goal together, closer to a real meeting than a single assistant answering one prompt.
If your bots' descriptions are specific enough, you don't have to explicitly mention who should pick up a task, name the job clearly enough and delegation becomes implicit.
Use a reply-to-isolate action on ambiguous follow-ups like 'okay, run that' so the agent doesn't misapply a short response to the wrong earlier message.

07Controls, Usage & Monitoring

Decide your default execution permission, always allow versus allow once, deliberately, it's the control that determines how much an agent can do on your computer without asking first.
Reserve human takeover for anything you wouldn't want sitting in plain text in a conversation log, passwords and API keys included, and rotate anything you accidentally paste in.
A shared group chat burns through weekly usage faster than a one-on-one conversation, so watch usage specifically once you start running multi-agent conversations.

08Managing Your Bots

Turn on notifications only for the handful of agents you actually need to hear from immediately, not every bot you've created.
Duplicating a bot is the reset button: it clears conversation history while keeping the skills and routines you already built, instead of losing everything.

09Agent Logging

Agent logging is what keeps a growing team of autonomous agents from becoming a black box, log what was asked, what was done, and what's still pending somewhere outside the chat itself.
Treat logging as a system instruction for every agent, not a one-off report: tell them to log at project creation and again at the end of every routine run.

Glossary

Terms worth knowing.

Grok Bot: xAI's multi-agent platform where each bot is a separate AI agent with its own name, description, and memory.
Plugin: A connector that lets a Grok bot sign into and act inside an external account or app, like ClickUp, Gmail, or GitHub.
Skill: A saved, editable recipe for a repeatable task that a bot can run on command or on a schedule.
Routine: An automation that fires a skill either on a time-based schedule or in response to an event like a webhook.
Webhook: A URL that, when an external system sends it a payload such as a form submission, wakes up a specific bot to act on it.
Compaction: Automatic summarizing of a bot's older conversation history so it doesn't hit its context window limit.
Handoff: One agent delegating part of a task directly to another agent, who then works on it and reports back.
Human takeover: A pause state where a bot hands control back to the person for sensitive steps it shouldn't handle alone, like entering a password.

Resources

Things they pointed at.

07:03channelAI Automation Society (Skool community) ↗

09:16toolClickUp

21:22productFree 2-hour Grok Bot course ↗

Quotables

Lines you could clip.

00:00

“I'm about to break down every single Grok bot concept that you have to understand in order to start getting immediate value out of your Grok bots.”

states the entire video's scope in one line→ TikTok hook↗ Tweet quote

12:17

“I know webhooks sounds like a technical word, but it's really not. Just think of it like someone ringing on your doorbell.”

turns a technical term into a one-line analogy→ IG reel cold open↗ Tweet quote

09:01

“Let's say you wanted to make chocolate chip pancakes. Well, you'd probably pull up a chocolate chip pancake recipe, then you'd make the pancakes... these skills are editable, and every time you run them, they can get better because you just give feedback.”

clean recipe analogy for an abstract product feature→ newsletter pull-quote↗ Tweet quote

21:22

“If you want to keep diving in to more stuff with Grokbot, then check out this free two-hour course that I just dropped.”

exact soft-CTA wording→ newsletter pull-quote↗ Tweet quote

The Script

Word for word.

Read-along

Don't just watch it. Burn it in.

See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.

metaphoranalogy

I'm about to break down every single Grok bot concept that you have to understand in order to start getting immediate value out of your Grok bots. So let's not waste any time and just get straight into today's video. So in today's video, I've got 25 concepts for you guys and they're split across five parts.

So part one is what the bot is. So a Grok bot is one single bot. As you can see, I've got a bunch over here and each one has a name, a job title, and a description.

So as you can see, if I click on Motion, which is actually building the animations for this video that you saw already, his name is Motion. You can see right here he's got a tag that says Animator. So if I click in, you can see that it says Label.

And the description for this bot is that this bot creates motion graphics and animations for intros or transitions for YouTube videos. And you can see that each of my different bots I'm having one very specific conversation with. So each bot is very much their own individual bot with their own skills, their own descriptions, their own processes, and you have unique conversations with them.

And now that's a really nice segue into number two, which is bot descriptions. Because you can see here that you're going to start to build out a ton of different bots, you want to make sure that each one is doing one very specific job very well. So in this case, you saw Motion's description.

If I went to Miner and we looked at his description, mine's X for AI news, new models, and new tools. Also knows Nate's YouTube and X posts, so the briefings stay useful for his content. If I go to this bot called Eyes, I can click in and see that this description is that this bot is a master at research.

It gathers lots of credible sources and helps form opinions based on facts and data. Now, the reason why these descriptions are so important is because as we get into different concepts later, you're going to notice that these bots work together a lot. And the way that they're able to understand, I need to delegate work to Eyes because he's the researcher, or I need to delegate work to Motion because he is the animator, is all based on those descriptions.

So these descriptions are super, super important. Because what's really cool about Grokbot is as you build more and you scale them up, you might think that you want to build some sort of ecosystem like this where you're up top talking to all your bots.

But really what you want to do is you want to build more of an ecosystem like this that looks more like an org chart where you talk to just a few executives and those executives know all of the bots underneath and they know which ones are on their teams and what their strengths and weaknesses are. And that's kind of what you can see what I've got going on up here with my C -suite, but we'll talk more about that later.

And that brings us on to number three, which is bot templates. So this is really cool. Do you remember how in NNN you could bring in other people's NNN templates?

In Claude, you can bring in... other people's skills, things like that. Well, in Grokbot, you can bring in other people's Grokbots.

So look at this link, for example. This is a Grokbot that I called Studio. It creates photos, graphics, and short video creative for marketing.

So what I can do is I can actually add this to my own Grokbot. It opens it up in the app. and I can read through the context here, so instructions, memories, skills, and then I can just add this bot.

And what happens is it spins up this bot, and then it just basically starts creating itself. Here you can see it says on it, I'm saving your memory and skill now. If I click into the bot, you can see it already has a name, a label, and a description.

Of course, you will have to configure your own things, so if someone gives you a bot that's connected to their ClickUp, it's not gonna be connected to their ClickUp, you would have to connect it to yours, but still, it basically has the other assets in there. just nothing personal. And the other cool thing is you can share your own bots.

So let's say I wanted to share my views bot. I could just come in here, click on the settings, and right here, click share as template. And this basically says, hey, views, create a template of yourself that I can give to someone else.

And there you go. Now views has created its own link, which I can go ahead and publish. And now if anyone copies this link, they'll be able to pull in views and have that bot in their own bot ecosystem.

Look at this. I'll copy the link. I'll go to Google, paste that in.

And now anyone can go ahead here and copy views. Which brings us on to number four, which is kind of a two -parter, context and memory. Context is basically all of the stuff that your Grok bots have access to to see.

So the conversation history and anything else that you're feeding in like your prompts. And the second piece of that is memory. And this is really interesting because right here you can see that when I asked views to make a template of itself, it said, yeah, I'm just gonna read through my memories first.

And if I come into the details, we can see here in the context that there are specific memories that it actually saved. And if someone copied this bot template, they would also get those memories. But what's really interesting is the difference between all of your GrokBot like global memory and individual bot memory.

So watch this. Hey views, can you quickly tell me what is the difference between what you know about me from a global perspective that all other GrokBots know and then what's the memory stuff that only you know about me? So the point I'm trying to make here is you have all these different bots and overall they have like a collective file of information about you.

But as you converse with each bot and you build up their capabilities, they have different memories individually too. Here you can see there's two distinct memory stores. This is what every bot is able to see about me.

And this is what only views knows about me. So make sure you are understanding what agents know what. And if you wanna update the memory, just say, hey, add this to your memory or add this to the global memory.

And now that brings us on to concept number five, which is compaction. So if you guys have used something like Cloud Code before, then you're probably aware of the whole idea of like filling up the context window. and being careful to make sure you're not entering into context -wrought territory.

Now with Grokbot, it's a little bit different because each individual agent has their own conversation history, which means they have their own context windows. But the thing about Grokbot is it auto -compacts everything. So we don't get to control that.

We also don't know exactly when it auto -compacts, but it has some sort of like auto -summarization and compacting capabilities so that it doesn't hit that context window. So if I've been talking to views for like 500 days, it probably wouldn't see my first couple of messages, but overall it would have memories kept about me and it would have - have context of what we're trying to work towards, and it would see maybe the last 30 or maybe the last 50 messages as a summary while it's taking action.

All right, so those are the first five. Now we're moving into part two, which is where it works. All right, cool.

So number six is the agent computer. Now what this means is that your GrokBots can actually use a computer. As you can see, Klaus has a computer right here and I can look at his screen and he can also control his screen.

If I go over here to Becky, you can see that she's also got a computer open as well. Every single one of your GrokBots has a computer. Now this is kind of similar to memory because they all kind of have their individual screens that they're working on, but at the end of the day, they're all working on the same computer.

And you'll understand that a bit better when we move into concept number eight. But moving on to concept number seven, is the idea of a browser.

So inside of the computer, obviously I can open this up and I can drive this browser or Klaus can because that's the bot right here. But what happens in the browser is we can literally just use Google the same way we might want to if we were actually on our own Google Chrome or whatever it may be. And what's cool about that is it saves your login.

So I logged into school, as you can see, this is my account. And if I close out of this, it's going to stay logged in. So now Klaus is able to actually do things inside of my school community for me, or it could log into my GitHub or it could log into my LinkedIn.

It can log into things. pretty much once we save the credentials. So you would basically just come into the site you want, you would log in, you would enter your own username and password, and then your bot could just take control and do stuff in there for you.

And now moving on to number eight, we have the file explorer. So just like on my screen, you can see right here, I've got a local file explorer. I can look through my Herc2 project.

I can look through my camera roll. I can look through all of that kind of stuff. So can all of your agents.

And this is where they share all of the stuff. If I open up our actual workspace, you can see that I've got stuff like brainstorms. I've got dashboards.

I've got Grokbot concepts, which is the video that you're watching right now. And all of these files that are being looked at. are the exact same ones on every computer for the bots.

So if I go over here to, let's just go to Dev, and I'll open up this screen, you can see that if we open up the File Explorer on Dev's computer, it's the exact same workspace with all of the same files that are stored across the shared computer for your GrokBots. And while we're still on this whole concept of the computer, moving on to number nine is the terminal.

So just like you have a terminal on your computer, so do these bots. They can run commands, they can set up scripts, they can do things through a terminal just like this as well. So it's literally just a computer.

They can do everything that a computer can do. And then moving on to number 10, we have local file access. So that means that yes, they have a cloud computer that they all share, but whatever computer that you're actually working on, they can access local files.

So right here, you can see that I gave it a file path, which was in my downloads. I said, analyze this file and tell me what's in it. It grabbed it, it analyzed it.

I told it to rename it and then move it. And it actually was able to rename it and move it. So if I go into my new thumbs, this is the file that it was talking about right here.

It used to be in my downloads and it was just called like, you know, orange, blah, blah, blah. And then it looked at it, it renamed it to zero to pro, and then it moved it inside of this new thumbs folder. And it did this locally on my own desktop.

All right, so those are the first 10. Now let's move on to part three, where we are going to talk about its abilities. Okay, so concept number 11 here is called plugins.

This is basically how you let your agents actually talk to things like your Slack or your Gmail or your Google Docs, whatever it may be. And you go down here to the plugins and you can see that there are a bunch of options to choose from and pretty much all you have to do is click add and then you just sign in. It couldn't be easier.

And what's also cool is you can see the ones that you're already authenticated into and you can connect multiple accounts. So let's say for ClickUp, I like to have my main account as well as a different account. So I can authenticate into both of them and then have different agents use different.

accounts so right here in my click up you can see that it's actually creating tasks for everything that it's doing here is when it started working on the motion graphics for this video and my grokbots know that they need to be logging their tasks inside of this board and click up so i can always see what's going on so once again this is the one for this video and you can see that we have an owner we have what i asked we have the deliverable we have links we have all of this information and progress updates and this is really good for me to just be able to keep tabs on what's going on and you know look back at the week and see everything that they did And moving on to number 12, this is skills.

So just the same way that you use skills in Cloud or Codex, you can do a slash command and then you could reference some sort of skill. So if I come in here and I type school, you can see that this one says like seven day challenge posts on school. And that is a skill that I could actually invoke and the agent understands how to use it.

And if you've never heard of a skill before, real quick, it's basically just a recipe. So let's say you wanted to make chocolate chip pancakes. Well, you'd probably pull up a chocolate chip pancake recipe, then you'd make the pancakes, and then you'd realize, okay, cool.

These were really, really good. I want to make them really, really good every time. So I'm just going to save this recipe.

So next time I want to make chocolate chip pancakes, I just read the recipe. And hey, if these weren't good, if they were too raw or they were too burnt, why don't I just update the recipe and say, hey, next time cook these a little longer on each side or cook them less. So these skills are editable.

And every time you run them, they can get better because you just give feedback. You find the skills inside of your plugins. You go to the installed and then you look at these private plugins.

And these are all of your skills that your Grokpots can use. So you can see I have like an AIS Plus recording post. I have Grill Me.

I have Fireflies and Jest. These are basically just processes that your agents will probably do more than just once, then you turn it into a skill. And a good skill has things like when to use it, the exact order of operations, and anything that if it could go wrong, what you do.

And speaking of skills, let's move on to number 13, which is teach a task. Whenever you're somewhere in the browser, I typically think that you do this on Chrome, but whenever you're in here, you can click on this button that says teach a task. And remember that skill that you guys saw earlier, the like seven day challenge post in AIS Plus?

This was a teach a task. where I basically came into here, I clicked on record, now you can see the screen's being recorded. It says Klaus is watching and learning.

So then I could click into this tab, I could go to news, and then I could go ahead and open up this link. And then when I go ahead and stop recording, Klaus gets that message, it watches that demonstration, and it turns what you just did into a skill. So this is going to be some skill called like find Chicago Cubs food news and whatever.

But the point being, you can either create skills with natural language and instructing what a process looks like, or you can basically have them shadow you. You could record for a couple minutes, pull some reports, analyze some data, whatever you need to do, and then it will automatically turn that into a skill for you.

As you can see, it decided to save that skill as Search Google News. Now, obviously, that's a pretty useless skill, but just trying to... demonstrate the point here.

And moving on to number 14, we have routines. You can see that right here, I have one called Sunday Work Log Archive, which runs every Sunday at 8 p .m. And in order to create one, you can basically just say, hey, can you do this for me every day at 5 p .m.?

Or you could just click on this plus and then set one up. Right here, you can see I'm setting one up called AIS 7 -Day Challenge Likes. And what I'm doing is I'm telling it to run a skill.

So routines make something happen automatically, and that thing that happens automatically can be a skill. So I chose interval, and I decided to let this one run every single minute. So I'll just sit here and let this trigger.

so you guys can see how that looks. So it's been a minute. You can see that this routine's firing.

We can see that Klaus is working down here. So it found the skill file, it's reading it, and now it's gonna be able to take action using that skill. And you can see in the browser, it just opened up a new tab.

I'll make this full screen and you guys can just watch it actually take action here and start liking those posts. So the top of the screen, you can see that it searched for the word challenge. It just clicked into a post and now it's scrolling down and it's going to click on the like button.

Boom, there we go. Okay, while that skill is finishing up, I'm going to move on to number 15, which is webhooks. Now, I know webhooks sounds like a technical word, but it's really not.

Just think of it like someone ringing on your doorbell. Because in the routine section, you can see that they can be triggered based off a time, like a schedule every hour, 5 p .m., or it can be off of an event. So these are event -triggered routines, like someone ringing your doorbell.

That could be on a Slack message, on a team message, or a webhook. And these are all basically the same thing. This one is just like you can configure your own webhook call.

So let me show you an example. I've got this bot down here called Webhook, and I created it just for this demonstration. If I go into this app that I created, what happens is I fill in this information, so I will just put Nate, I will put nateherk88 at gmail .com, and I will put hello, I need a new dog.

And when I send this request, what this does is it actually sends a webhook. So it basically just shot a payload to this webhook address, and you can see now this bot has woken up, it's going to take that payload, so basically the information I submitted in that form, and it's going to take action for me.

So you can have agents automatically wake up when you get a new email or a new Slack message or a form submission or so many other possibilities, a new CRM record even. So what happened is it said, hey, a form just came in asking for a new dog. I emailed them that we got it and we're processing it and I DM'd you in ClickUp.

So if I go to ClickUp, you can see that I got a message from UpItAI. If I can make this a little bigger, you can see that that form just came in. And in my email, you can see that we just got a message which says we got your form and you wrote, hello, I need a new dog.

We'll follow up soon. So obviously just a quick demonstration, but the point being, you can have GrokBots take action when something happens in your business or in your life. Okay, so that was 15.

Now let's move on to part four, which is how bots work together. This work gets like really cool. Okay.

So the first one is group chats. Now look at this. I have a group chat with my leadership team.

I've got Klaus, Dan, Chandler, Becky, and I could add members if I wanted to, or I could remove them. And in here, I basically said, Hey, I need you guys to like, tell me about our long -term goals and ideate on how we're going to keep growing and scaling the business. And because these are different agents, my chief of staff, COO, CFO, CCO, they're able to talk about their own areas of expertise and all the individual agents that they're actually managing.

Remember, I like to use sort of like this vertical hierarchy and they're able to to all hear each other, mention each other, and just ideate, and then I can come back and take a look at what they've done. And that brings us on to number 17 really perfectly, which is mentions.

You can see that Klaus mentioned Becky right here. You can see that Dan mentioned Becky and Klaus right here. And if I click on that, it actually takes me to that conversation.

So you can actually see too that Klaus just finished that skill. And I'm going to go ahead and disable that routine because I don't want that to actually fire every single minute. But what else you can do is you can have agents delegate work out by mentioning people.

So like, hey, could you please check with, and then I could just go in here and mention, you know, like at Becky to do X, Y, and Z. But the cool thing is if your descriptions are good enough, you don't actually have to explicitly mention these bots. Klaus or whoever you're talking to will be smart enough to know that it should be delegating work, but you can still message them in chats just like that.

And that brings us to number 18, which is about handoffs. Like I said, agents delegating work to other agents. If I go to Becky right here, you can see that I asked her to do some stuff for creating these motion graphics that you're looking at.

And she delegated all this work to motion. You can see right here. Now in here, if I click into motion, we can see that the actual conversation between them.

So Becky said, hey, Nate needs this. Here's what you need to do. Here are the deliverables.

And then they have a conversation. And it can also be more than just one agent. Right here, Klaus was able to delegate work to...

and Minor. And then down here, it also shot off messages to Minor and Becky. The whole point is agents can work together and you can also view their conversations so you get full visibility into what they're actually doing and talking about.

Which brings us on to number 19, which is about replies. Say you're talking to Klaus and there's a bunch of things going on. A routine just fired, you were building a skill, whatever.

You can use this little reply button to actually isolate that conversation sort of. So if I click on this and say, okay, very cool. Go ahead and run that as a test.

Nice. And then I can also respond to this, for example, and say, very nice, well done, you're crushing it. And just like the way you reply to things when you're actually in Slack or ClickUp, it's just able to understand what you're responding to and the context is a little bit better.

Because if I randomly just said, okay, cool, go ahead and run that as a test, it might think that I wanted it to do this again, which I, in fact, do not. Okay, so we've been through 19 concepts. We've got six more to go.

Moving on to part five, we have how you control and monitor it. So the first one is about approvals. If you go in here to your settings, you can see that we have computer.

And this shows execution on this computer. Let Grokbot open files and run tasks on your computer. So this is my actual local environment.

Remember we talked about the difference between local files and the sort of like agent computer in the cloud. Now, sometimes when you're talking to these agents, they're gonna ask you, they're gonna pause the workflow and say, hey, am I allowed to do this? And you can either choose like always allow or allow once, and you can change your default setting over here as well.

Which brings us into concept number 21, which is human takeover. Sometimes it's gonna hit a blocker where you're really the only one that can take over. For example, if you're asking it to do something in LinkedIn and it says, okay, I've pulled up LinkedIn in my browser, but you need to sign in.

It will pause and it will wait for human takeover. This is usually when you need to do something sensitive like put in passwords or anything. sort of information like that because what you don't want to do is put passwords into the conversation history right here because technically that's going to spacex i'm not saying that anything really really bad would happen but if you do accidentally send an api key or password then maybe just go ahead and rotate it and moving on to number 22 we have weekly usage so right here when you sign up for grokbot you get on some sort of subscription whether it's like 20 bucks a month or 100 or 200 whatever it is you can see weekly usage and this resets every seven days so if this goes to 100 within your seven days then you can't use this anymore unless you wanna pay extra for on -demand spend or API billing.

And you can just click on this button if you wanna change your limit, increase it, decrease it, or add some on -demand credits. So as you're dealing with different bots and as you're sending off prompts and you're making different channels, just watch this number because having things like a group chat of agents is going to consume more of your weekly usage than just talking to one agent.

And moving on to number 23, we have run history. This basically just means the ability for you to check. how many times routines have been run.

So for example, with Webhook, this one is a event -based one. So in the run history, we can see that we had one 10 minutes ago and we also had one at 5 .48 PM. Or if you have something that's actually time -based, you could just click in to verify that they are in fact running every Sunday at the time that they should be.

Or if this one, you know, every minute, this would have tons more history. Now, unfortunately, you can't click into them. You can just see that things ran, but if they failed, there'd be an X here.

Hopefully in the future though, we're able to actually click into the runs and see where they actually happened in case we need to investigate what went wrong. I do kind of have a fix for that though, so just stay tuned. And moving on to number 24, we've got notifications.

Now this one's pretty self -explanatory. It'll notify you if it has a blocker or if something finishes up, but you can also control notifications per agent. So what I like to do is just have notifications on for my main four and you can come in here and you can just enable notifications when the bot finishes or needs input.

Now what's really cool about that is that you can also have the notifications come to your phone, which is what I like. You can see right here, we've got my phone and it's perfectly synced with my desktop. So if I open up Klaus and on my phone, I just say hello.

As soon as I send this, it pops open on my desktop as well. So even if my computer's turned off, these both stay perfectly synced. And getting the notifications on your phone is awesome because while you're at lunch or on a walk, you can keep driving what you need to drive from your pocket.

And finally, the last one we have is number 25, which is basically just about managing your bots. Just the idea that you can right -click on them and you can pin them, which moves them up here to the top. You can also move them to a different section.

So if I unpin this one and then I go to move to, I can move it to a new folder called... whatever I want, or I could move it to the general or unassigned, or I could put it into wherever I want, really. And this allows you to just fold them up and keep things organized if you choose.

And you can also hide bots. So for example, I'm going to delete Webhook, but if I don't want to delete him, I could just hide from sidebar. And that just puts him in a section down here called hidden bots.

So these aren't lost, but they're just hidden. So you can keep things clean. You also have the ability to delete them.

You have the ability to share them as a template. You can edit their profile. You can mark them as unread.

And you can also duplicate them, which is interesting. because it basically just spins up the exact same bot, but it clears all the conversation history. So let's say for some reason you're getting sick of how Klaus is behaving and maybe you're like, oh, I just need to restart, but I don't want to lose all the skills and all the routines and everything.

I just want to like clear the conversation history. The way you would do that is you would just make a duplicate copy of Klaus and then maybe just delete the old one. And then the last one here is just to copy the conversation ID.

This is really just more for if you need like support and you want to send, you know, Grok your conversation ID so it can look into. What happened? And that was all 25.

The last one I have real quick for you guys that you kind of already saw a little teaser about was number 26, agent logging. Just the idea that everything that these agents are doing, you want to store them somewhere. Whether that's a Google Sheet, whether that's a ClickUp like I've got, whether that's a local Excel file or a SQL database, whatever it is, store everything your agents do.

meaning tell them whenever you are creating a new project, log the updates there. Updates, progress, move it to complete, archive it, things like that. And then what you can do is say, hey, at the end of every single routine, you're also going to log this run.

And now you can have different databases or different sections to just see every single thing that your agents are doing. And I think that's really important when you're using something like Rockbot, where it's so easy to make autonomous agents and it's so easy to just scale up a team. You don't want to lose visibility and you don't want to lose control.

So please set up some form of agent logging. But anyways, guys, that is going to do it for today. If you want to keep diving in to more stuff with Grokbot, then check out this free two -hour course that I just dropped.

Go ahead. I'll see you guys over there. Thanks for watching the video.

The Hook

The bait, then the rug-pull.

Nate Herk opens with a promise to cover all 25 Grok Bot concepts you need before building a team of autonomous agents, then spends the next twenty minutes proving it, concept by concept, inside the actual product.

Frameworks