Grok Bot Full Tutorial for Beginners: How to Use & Set up Grok Bot

0:00Every AI tool that you're using right now dies the second you stop talking to it. Ask a question, get an answer, and then walk away, and the work just stops. Nobody tells you that when they sell you on AI automation. And in this Grok Bot full tutorial for beginners, well, I'm going to go ahead and show you how to break that pattern completely. Because for most people, AI still starts and ends with, well, a chat box. You prompt it, it replies, you prompt it again, etc. And that's it. But that isn't automation. That's just a faster conversation. Now, the real shift happens when an AI stops waiting for you

0:37and then starts acting like something with an actual job, role, right tools, a process that it can repeat on its own, and a schedule that it runs without you having to sit there babysitting it. So, that is exactly what I'm setting up in this video. I'm going to walk you through how to use and set up Grok Bot step by step, how to connect it to real tools so that it can actually move work forward instead of just talking about it, and get it running like a digital worker instead of just a chat window you have to keep refreshing. So, that by the end of this video, you're going to see

1:10exactly how far we can push this, an AI that knows its job, that uses real tools, and that keeps working even when you're not there telling it what to do next. Now, if you do want to master AI tools and you really do want to learn how to build profitable SaaS apps and websites and AI agents and mobile apps with AI, of course, I've created my own complete master class that shows you exactly how to do all of this step by step. Now, my master class normally costs $499 to join, but since you are taking the time to watch this video, and I appreciate you, you can join

1:41completely free. Just check out the link in the description to get free access to my master class, and you too can start building your own AI-powered business today. So, let's go ahead and start up and get into this. Before we build the bot, it does help to get the setup out of the way and really understand what Grok Bot is actually running on. Now, the app itself may be installed on your computer, but the bot's real workspace lives up in the cloud, and that difference becomes important once we actually do start automating things a little bit later on. So, to get started here, let's head on over to x.ai/bot

2:16in your browser and click download in the top right corner. So, GrokBot is available for Windows, macOS, and iOS. So, choose the version that's proper for your machine and run the installer once it finishes downloading it. After that, just follow the normal installation steps that we're used to, accept the required agreements, and then just finish the setup to launch the app. Now, the native app becomes the sort of control interface that you use to communicate with the cloud environment where your bots will actually run. And the first time that GrokBot opens, click sign in and log in using an account with access to the platform. Rather than say

2:52dropping you into an empty workspace, by the way, GrokBot will start you with a short onboarding conversation. It asks what kind of work you mainly do, such as say maybe productivity, maybe engineering, or some design, or marketing, and which tools you normally work with, like Google Workspace or maybe Microsoft 365 or even Notion. So, your answers will help it understand what you might want assistance with, and then also lead you directly into creating your first bot. After onboarding, the main workspace opens, and the layout is fairly easy to understand here. The left sidebar contains your bot roster. So, every bot you create will appear right here. And

3:36you'll also find marketplace in the same area for managing plugins and other integrations, while your profile menu here at the bottom gives you access to your account settings and your subscription usage. The center of the screen here is, of course, the chat panel, and this is where we'll talk to the active bot, give it instructions, and review anything that it returns. And then on the right, the cloud computer shows the virtual environment that the bot is working inside of, including browser activity as it happens. And then directly below that is the routines area, which we'll use later to create scheduled automations and triggers. The cloud computer is worth paying attention

4:16to here because it does explain why Rockbot can work differently from say a normal desktop app. The bot itself is running on XAI's infrastructure rather than relying on your computer to stay awake. The desktop app is mainly just there to be the window that you use to control and watch it. Closing Rockbot, losing your internet connection, or even shutting down your computer does not stop an active or scheduled workflow running in the cloud. Now that the app is installed and we know where everything lives here, we can go ahead and create the weekly research bot and start giving it an actual job. Creating a bot gets a lot easier when you stop

4:53thinking of it like say starting a new chat and start thinking of it more like assigning someone a human a role. And before it can do useful work, it does need to know what its job is of course, how it should communicate with you, and what boundaries it should follow. All right, so to create our first one, let's click the plus button here in the top left corner of the sidebar, and then just choose create new bot. And for this build, let's ask Rockbot to create a weekly research bot with the job title weekly intelligence analyst. Now, its role will be to research assigned topics online, to use verified information, to

5:30write concise summaries, and then send them through email. We also want it to communicate in a direct executive tone, to avoid unverified or fabricated information, and to never make financial transactions or change account settings without explicit permission. Let's send that instruction over and Grok Bot will then create the new bot and then add it to the bot roster here in the left sidebar. Click it there to open up its workspace. The job description is one of the most important part of the setup here because Grok Bot keeps referring back to it as it works. So, when it runs into unclear information, maybe unfamiliar tasks or situations where it

6:13needs to decide what it is allowed to do, then these instructions help guide that decision. So, a clear role also keeps its writing style consistent and it gives it boundaries to follow during future automated runs. Before connecting any plugins or scheduling anything, it is worth checking that the bot actually understood those instructions. So, here in the chat, let's ask it to tell us its job title, to explain its main responsibilities, and to describe how it handles unverified sources. And then the responses should confirm that it is the weekly intelligence analyst, that its job is to research topic, that it gathers verified information, that it writes concise reports, and it sends

6:56them by email. It should also make clear that it will not invent missing details or rely on questionable claims, and that anything it cannot verify will either be flagged or just completely left out. So, that quick test then gives us a simple way to confirm that the bot understands its role before we even start giving it access to external tools and automating the workflow here. A research bot can also do a lot on its own, but the moment that we want it to actually send something outside Grok Bot, it does need access to the right tool. Now, since our weekly research bot is supposed to email

7:31the finished report, then Gmail is the first plugin that we obviously need to connect. And we can do that directly from the chat itself rather than having to dig through a separate settings menu. So, let's tell the bot next that it will be sending and delivering the summaries through Gmail. And as soon as we send over that instruction, Grok bot will then recognize that Gmail access is required and brings up a connect Gmail card inside of our conversation. So, of course, let's click connect, choose the Google account that you want to use, review the permissions, and then click continue to authorize access. And after the OAuth process finishes, return to

8:11Grok bot again, and then chat should confirm now that Gmail is indeed connected and it's ready to use for sending those reports. The useful part here is that plugins are connected at the account level rather than being locked to just one bot. And since we have connected Gmail here, any other bots that we create later on can also use that very same integration without making us sign in all over again. The same idea applies to tools like Google Drive and GitHub and other supported services. You just connect them just once, and then they become available across your entire bot roster. So, that shared setup now becomes especially

8:47useful as you start building multiple bots that need to work across the same apps rather than say having to reconnect everything for every new workflow. Now, repeating the same detailed instructions every single time will get old quickly, especially if this bot is supposed to run on its own every week. A skill solves that by saving the process just once so that the bot already knows how to handle the task whenever it does come up again. So, for our weekly research bot here, we need a research and report skill. And that defines exactly how each briefing should be created. So, let's ask it to build that skill, and then

9:27tell it to research the assigned topic using reputable sources from the say past 7 days, and then turn the findings into a 3 to 400-word report. The finished report should include a headline, a key takeaway section with three concise points, a short conclusion, and a verified sources section containing the direct URLs, all formatted cleanly in markdown. So, send that instruction over and let Grok Bot do its thing and it'll build the skill and register it automatically. So, there's no separate save button that we need to click here. Once it finishes, the research and report skill should just appear among the bots' saved skills. We can also check it from the

10:09marketplace itself in the lower part of the left sidebar here. So, let's open up marketplace, go out into the library of created skills, and then installed apps, and we should see the research and report skill listed there. Now, opening up the skill lets us also review its instructions to change how it works maybe later on or to just delete it if we no longer need it. What's important is that the process is now saved rather

10:35than tied to just one individual prompt. We can give the bot a completely different research topic later on and it will still know how to verify recent sources, to structure the report, to keep it within the right length, and include the source links without us having to explain those rules all over again. And because skills live in the shared workspace, they can also be reused by other bots that we also create later on. Now, a bot is much more useful when you don't have to remember to start it yourself every single time. Now, since this report is supposed to arrive every Monday, then the next step is

11:12giving Grok Bot a schedule so that the entire process can happen automatically in the background. And that is what routines are for. A routine tells the bot when to run and what task to carry out, whether that means following a fixed schedule or maybe just responding to another the And for this build, we want the weekly research bot to run every single Monday morning without us having to open up the app or send a message to it first. Let's ask Rock bot then to create a routine called weekly Monday research briefing that runs every Monday at 9:00 a.m. And each week it should research the latest developments

11:52in autonomous AI agents, use the research and report skill again that we just created, and then email to us the finished summary. So, let's go ahead and again send over that instruction, then open up the routines tab here in the right side panel below the cloud computer. So, our new weekly Monday research briefing should now appear there with Monday at 9:00 a.m. as its scheduled time. Rock bot has taken the natural language instruction and then connected the schedule to the saved skill and then added the email delivery step for us without making us configure each part manually. The routine card should also show that it is indeed

12:31active. So, opening it up, let's just review the full setup to maybe say change the schedule later or pause it if we ever want to stop the automation temporarily. This is also where the cloud setup that we saw earlier starts to become genuinely useful because the schedule is running on XAI's infrastructure. So, your laptop, it doesn't need to stay open and the Grok bot app does not need to be running either. Even if say your computer is turned off or you're just away from it, the bot can still wake up on Monday morning, research the topic, create the report, and send the email to you on

13:08schedule. Building the workflow, it's one thing, but the part that really proves everything is working is watching it just complete the entire job on its own. Now, we already have the role, the Gmail connection, the saved skill, and Monday routine in place. So, now we can just run the full process and see whether all of those pieces actually work together from start to finish. Open up the routines tab here on the right side panel and then select the weekly Monday research briefing. Now, we obviously do not need to wait until next Monday at 9:00 a.m. to test it out. So, click test run to trigger the routine

13:44immediately. And then the button changes to running, which does confirm that the workflow has started on XAI's cloud infrastructure. From here, the bot handles the entire process in the background. It searches the web, it checks the sources, it writes the summary, it formats that email, and then it sends it without filling the screen with every single individual action or requiring us to monitor it along the way. And after the run finishes, let's look back at the central chat panel here. A completion card should appear automatically confirming that the weekly research briefing was indeed successfully sent to the connected email address. It also shows the generated subject line, something like weekly

14:26research briefing autonomous AI agents, along with a short overview of the main topics that it found from reputable sources that were published during the past 7 days. So, that completion card also gives us confirmation that the routine did run from the beginning to end and successfully deliver the result. Now, let's open up the inbox and

14:49actually check what actually arrived. Let's find that email using the research briefing subject line and open it up. So, here at the top, we should have a clear headline summarizing the topic, followed then by an overview covering the main developments, the technical updates, and the industry events the bot found during its research. And then as we scroll further down and then now here, we can check whether it followed

15:15the research and report skill correctly. The key takeaways section should contain exactly three concise points followed by a conclusion that does bring its findings together and explains the more broader implications. And then here at the bottom, the verified sources section should include direct clickable URLs for every report or news item used in the briefing. So now the bot has followed our saved SOP, it formatted the report clearly in markdown, and it delivered a finished briefing without us having to manually research, write, format, or send anything. For someone new to AI agency, all of this happen just makes

15:57the idea much easier to understand. Chatting with an AI can still feel like using a more capable search or writing tool because well, you are constantly giving it the next instruction. But here, the bot completes a real workflow and delivers a result to an inbox through a process that can run on a schedule without someone having a prompt each time. And that is the kind of autonomy that we're actually trying to build here. There is also another layer that we can add as well. Grok bots can pass work to other bots, which then means one automation does not necessarily have to handle every single part of a larger workflow by itself. I

16:38already have another bot here called social media bot and its job is to turn research summaries into posts for X and LinkedIn. So let's connect it to our weekly research bot and then see how that kind of handoff works out. So open up the weekly research bot here again and then let's tell it that after finishing each report, it should pass the summary to the social media bot so

17:03that it can create a short update post. Use the @ symbol to bring up the bot selector, choose social media bot so that it appears as an active mention, and then send off the instruction. The weekly research bot should now confirm that it will now hand over each completed research summary to the social media bot after every run. It also updates the existing routine automatically. So, let's open up the routines tab now again, select weekly Monday research briefing, and check that

17:33the action steps are actually there. That the fourth step should now include an instruction to message social media bot with the finished report once the research workflow is complete. So, now let's test the updated version right away. Let's click test run again, and then watch the central chat panel here as the weekly research bot goes through its research cycle. As it finishes the briefing, it sends the email, and then passes over the completed report directly to our social media bot. And then after that handoff, let's look at the bot roster here on the left. An unread notification badge should now appear beside our social media bot. So,

18:13let's open it, and here we can see an incoming card labeled message from weekly research bot containing the research summary that was just generated. Then the social media bot can take that information and create content based on its own specialized role. For X, it produces a shorter, punchier update covering things like the latest Asia releases, managed runtimes, and

18:38notable safety evaluation incidents. Now, the LinkedIn version is a little bit more structured and executive focused covering topics such as platform acceleration, potential cost savings, and enterprise security requirements. So, now we have two specialized bots working as part of the same workflow. The weekly research bot handles the research, the verification, the briefing, and email delivery, and then it passes the result to another bot that then turns the same information into platform-specific social content. Grok Bot's internal messaging systems allow these hand-offs to happen automatically, which then opens the door to building even larger multi-agent workflows where different bots each can handle the part

19:25that they are actually best suited for. So, after putting this much work into a bot, there is a good chance you will need to and want to reuse the setup somewhere else or maybe to share it with someone without making them rebuild everything from scratch. Thankfully, Grok Bot handles that through templates, which then package the structure of the bot while keeping your personal accounts and private data still separate. So, to create one, click share in the top right corner here of the chat panel, and then just choose share as template. Grok Bot automatically starts the template creation process within the chat, essentially asking the bot to package a

20:04shareable version of itself. And then, as it builds the template, Grok Bot reviews the parts that really make up the weekly research bot and then shows you what will be included. And that covers its memories and preferences such as the delivery rules, the 7-day verified research format, the briefing details, and the multi-agent hand-off that we added earlier. It also includes the research and report skill, the weekly Monday research briefing routine, and of course, the requirement to connect to Gmail. Now, before anything is shared, Grok Bot applies a privacy filter. So, your personal details, such as private conversation history, your email addresses, internal account IDs, OAuth tokens, API keys, and other

20:51account credentials, those are removed from the exported version. The template just keeps the structure of the workflow without giving someone access to your actual accounts or your private data. And then, once the template is ready, a card will appear inside of the chat showing the bot icon, the weekly research bot name, its description, and its current status. You can go ahead and click view details first to inspect the full setup and really make sure that everything looks right before sending it off and sharing it. And when you are ready, click publish to make the template available through a shared link. Then use copy link to copy that

21:30URL, and then anyone who opens it can go ahead and launch GrokBot and add their own copy of the weekly research bot to their own roster in just one step. What they then receive is basically a duplicate of the configuration rather than access to the bot that we've been using ourselves. Their copy includes the job description and the operating rules, the saved skills, the routine schedules, the required plugin setup, and the multi-agent delegation rules. They do not receive access to our live cloud computer, to our connected Gmail account, OAuth credentials, private chats, any account IDs, or any of our personal memory. So, that makes templates very useful for sharing

22:12complete workflows without sharing the sensitive accounts behind them. So, someone can go ahead and take this research bot, for example, then connect their own Gmail account, adjust the topic or the schedule, and then just start using the same structure inside of their own workspace. XAI is also developing a public marketplace for bot templates where now creators will be able to publish, showcase, and potentially monetize their configurations. And that makes experimenting with templates now especially useful for anyone who really wants to build polished workflows and start sharing them before that ecosystem, you know how it is, becomes more established. What I want you to remember here is that GrokBot is useful

22:57when all of the pieces are working together. The research itself is only one part of it because what really changes the entire experience is being able to give the bot a clear role to connect the right tools, save the way it should work, and then just let it run without needing anyone to remember every single step. And if you do want to go further, x.ai/bots/guides is the official guide library for learning more about the platform, for exploring additional plugins, and even building more advanced workflows. You can also experiment as well with shared bot templates if you would just rather start from something that someone else has already built then just adjust the

23:39skills, the routines, and the integrations to better fit your own use case. And then as you become more comfortable with that kind of structure, you can go ahead and start combining multiple specialized bots into just larger automated workflows. All right, so that's it for this one. I want to thank you for watching and investing your time with me here today. I'll see

23:59you at the next build.