When Grok Bot launched, I was skeptical about it.
I already use Claude Code, ChatGPT, and Codex. I did not want another AI tool asking me to rebuild my whole workflow from scratch. Dheeraj Sharma felt the same way. If something already works for him, he would rather improve it than jump to whatever launched two weeks ago.
But after using Grok Bot for a few weeks, I found a reason to keep it in my current setup.
I can create several bots, name them, give each one a specific job, connect them to the apps those jobs require, and access the same bots from my phone. All of my bots run in the cloud, so I don’t need to keep my laptop open or prompt them manually. I built a routine that lets these bots run without me.
By Episode 26 of One Shot Show, I had seven Bots sitting in the sidebar: a Social Content Writer, Chief of Staff, Invoice Hunter, Content Researcher, SEO Agent, Creative Director, and Business Coach.
During the session, I showed Dheeraj and the rest of the audience how I use them, what each one can do, and which apps these bots are connected to.
But first…
What Is Grok Bot and How Does It Work?
Grok Bot is an AI agent app where you create and name Bots and give them work to do across your apps, websites, and files. You describe the result you want in a conversation. With the necessary access, the Bot can research a topic, prepare a document, update a content backlog, or carry out a recurring task.
The software gives those Bots a cloud computer with a browser, file storage, and a terminal for running commands. That’s how a Bot can execute tasks inside other apps. When I ask for research to be added to Notion, the Bot can figure out how to get the information it needs and execute an authorized connection to the place where I want the output saved.
In my usual Claude Code or ChatGPT setup, I begin with a project folder and open conversations inside it. In Grok Bot, I open a conversation with the bot responsible for that job: it can be a content writer, a researcher, or my chief of staff.
What Can Grok Bot Do? 6 Core Features
Before we get into my seven Grok Bot agents, there are a few features that explain how that works.
1. How Do You Give a Grok Bot a Role and Memory?
You can create multiple Bots and give each one a name, description, and responsibility. The description explains the work it owns and the boundaries it should follow. Memories add information you want it to carry into future conversations, such as your audience, preferences, or business background.
For example, I gave my Content Researcher information about AI Maker and the kinds of ideas I find interesting. That gives it something specific to work from when I ask it to investigate a new topic. The name helps me find the right Bot; the information behind it helps the Bot understand the request.
2. Does Grok Bot Work When Your Laptop Is Off?
The bots can keep working when I close my laptop because their computer runs in the cloud. I do not have to keep my own laptop awake for those cloud tasks to continue.
All bots on an account share that computer, including its files, app logins, and signed-in browser sessions. But each bot can run separately, so you can request totally different things from different bots, even though at the core they are all running on the same shared cloud computer.
3. Which Apps Can Grok Bot Connect To?
Plugins and connectors let the Bots retrieve information or take actions in services such as Notion, Gmail, Drive, Granola, or even Typefully. You may already know these these connections as MCP, a standard that lets an AI agent use tools provided by another service.
But what’s interesting to me, compared to Claude Code or ChatGPT, is that to connect apps you can simply do it in the conversation, instead of clicking a button. You just sign in and authorize it when prompted. Grok Bot can also use its browser to work with websites. The available connections and the permissions you grant determine what it can actually do.
4. What Are Grok Bot Skills?
Same as Claude Code or ChatGPT, Grok Bot also has Skill.
A skill records how a task should be done. Once you have a process that produces a useful result, you can ask the Bot to save it for reuse. That process might describe how to research a topic, which sources to check, and how to format the findings.
This is where you can carry your preferences into repeated work. If you want every traffic report to compare the same periods and include the same charts, those instructions belong in the process the Bot follows.
5. How Do You Schedule Tasks With Grok Bot Routines?
A routine gives a Bot recurring work on a schedule. A routine can be effective if you combine them with Skills.
The skill describes how to prepare the report; the routine tells the Bot to prepare it every Monday morning.
Since Bot runs in the cloud, Routines no longer require your laptop to be open to execute tasks with tools like Claude and ChatGPT. You can run any scheduled task—for example, getting a morning digest from your email at 9 AM—even while you’re away from your laptop.
When setting up Routine, the official guidance recommends testing the process before scheduling it.
6. Can You Use Grok Bot From Your Phone?
The iPhone app connects to the same Bots and conversations I use on desktop. I can start a task, answer a question, review a result, or approve a step while I am away from my desk.
This is the combination I kept coming back to during the session. The Bot already has its job and background, its connections provide the information it needs, and the cloud computer lets the work continue after I put my phone away. Here is how I divided my own work across that setup.
7 Grok Bot Agent Ideas I Use Every Week
Each Bot owns a narrow part of my work. Some are already useful. A few are still early experiments:
1. Social Content Writer
The Social Content Writer starts with a finished newsletter or an idea. It can turn that source into a LinkedIn post, an X post, or a Substack Note, then use Typefully to schedule them all at once.
I showed one previous run where I attached an image, asked for a Substack Note with a specific caption, and had the Bot move it into Typefully. The Note appeared on my Substack account, thanks to recent Typefull and Substack integration.
During the live session, I also asked it to research and write a LinkedIn post about using Grok Bot to build multiple agents. I reviewed the draft, then asked the Bot to schedule it for the following day and add a newsletter call to action in the comments.
In my current setup, I taught the bot how to write for LinkedIn. If I share ideas, it first needs to do research to back up the claims and provide some facts before deciding to write the post.
Here’s an example of how I use it: I write and schedule my social posts from my Grok Bot iPhone app. I use Wisprflow to brain dump, the bot then researches the ideas using Tavily, and finally schedules the post using Typefully.
Here are the apps currently connected to my social content writer: Typefully and Tavily.
2. Chief of Staff
The Chief of Staff can read my calendar, email, and Todoist. The routine I described checks those sources in the morning and turns relevant commitments into a task list I can review on my phone.
The appeal is simple. By the time I open Todoist, things I need to do for the day are already there.
List of connected apps: Gmail, Calendar, Todoist.
3. Invoice Hunter
The Invoice Hunter has one specific job: find invoices in my email, pull out the relevant details, and organize them in a Google Sheet or report.
This is the kind of task I want an agent to handle because it’s an admin job I hate doing and it’s repetitive. So I outsource all of it to the bot.
List of connected apps: Gmail and Google Drive.
4. Content Researcher
The Content Researcher connects to the sources behind my content decisions and help me to manage my content backlogs. In the version I showed, that included YouTube data through Supadata, X, Notion, and Tavily.
If I have an idea for a newsletter, I can tell the Bot to research it across YouTube, Internet via Tavily, and X, then add the findings to my Notion content backlog. When I return to Notion, I have source material waiting for review instead of an empty idea card.
List of connected apps: Supadata, X, Notion, and Tavily.
5. SEO Agent
My SEO Agent uses OpenSEO alongside my Google Analytics and Search Console data. I use it for keyword research, competitor research, and recurring traffic reports.
I have a routine that checks the previous week’s performance every Monday morning and compares it with the week before. I also showed an HTML slide deck the Bot had created from my August 2026 analytics, including traffic channels and engagement data.
That output was useful because I could open the visual report directly inside Grok Bot.
My SEO agent can also run keyword research and competitive analysis when I want to target specific keywords before writing them in my newsletter. This helps me understand how to write more unique angles compared to what other people have already published on the internet.
List of connected apps: OpenSEO and Google Drive.
6. Creative Director
The Creative Director can use Grok Imagine to generate images. My plan is to teach it the visual rules I use for AI Maker so it can produce a closer first version.
I had only created this bot shortly before the session, so the design guidance was not ready yet. The image generation worked, but I still need to teach it my visual guide.
7. Business Coach
The Business Coach knows about my newsletter and can help me think through questions such as which lead magnet to build. I’ve modified my Business Coach into someone who disagrees with and challenges my ideas, not just supports them. I’ve learned interesting tools I can use to improve both my free and paid subscriber growth from it.
Whenever I have new ideas to explore, this is the first agent I go to.
How Do You Connect an App to Grok Bot?
I find that connecting apps using Grok Bot is much easier than with Claude or ChatGPT, especially for apps that don’t have official integrations in the Connectors or Plugins menus.
For example, when I wanted the Content Researcher to access X, I asked it to check my feed. The Bot noticed that it did not have the required connection and returned a button that started the authorization flow. I used a similar process with Notion.
This means that the Bot can find the right app connection from a simple chat, which saves me so much time from hunting MCP docs manually.
This feels more like an agent that actually helps me do the job instead of needing me to tell it how to do the job.
The current product documentation makes another distinction that I did not explain clearly during the stream. Installed plugins are available across the account, while the Bot’s role, memories, skills, and routines shape the job it performs.
Grok Bot vs Claude Code: What’s Different?
Grok Bot keeps the interface quiet while a bot works. I saw a simple animation and waited for the result. I didn’t get the same detailed view of files, commands, and intermediate actions that I’m used to seeing in tools such as Claude Code.
That simplicity will probably help someone who wants to describe a job and review the output without studying how the agent did it.
I felt the tradeoff during the live demo. The content research and LinkedIn scheduling tasks could continue while I showed other bots, but I had less visibility into why they were taking time. When the job is sensitive or the process itself needs debugging, I prefer seeing more of what the agent is doing.
Another thing I want to highlight is that there’s no model or thinking-capability selection in Grok Bot. I think that’s fine for me, considering I use the bot to automate tasks I don’t want to do manually. But for deeper work that requires huge token consumption and a local Markdown project folder, I still come back to Claude and ChatGPT.
Grok Bot currently fits the jobs I want to start from my phone, run on a schedule, or leave working in the background.
How to Set Up Your First Grok Bot
Seven Bots is where I ended up after experimenting. I would suggest you to start with one.
Pick a task that repeats and produces something you can check. Think about something you do every day: a weekly traffic report, an invoice list, creating and scheduling social posts, or a set of research findings is easier to evaluate.
Then define five things:
The job: What repeatable result should this Bot own?
The sources: Which apps or files contain the input?
The output: Where should the finished result appear?
The boundary: Which actions require your approval?
The schedule: Does this need a routine, or should you run it only when needed?
Run the job manually through the Bot first. Watch what it misses, correct the instructions, and turn it into a routine only after the result is stable enough that you can rely on overtime.
For me, the best first bot from this list is the Content Researcher. It pulls sources from Tavily, YouTube, or X, and sends the output to Notion, where I manually review the results later.
I’m still early with Grok Bot, but the live session gave me enough evidence to keep using it and pushing its capabilities.
That is a useful bar for a new tool to clear before it earns a place in my current AI stack.
And that’s how you can try them too.
Show Details
Show: One Shot Show
Episode: 26
Topic: Building and using a seven-Bot Grok Bot team
Hosts: Wyndo and Dheeraj Sharma
Live schedule: Wednesdays at 10:00 AM ET on Substack
Timestamps
00:00: Episode 26 introduction and why Dheeraj is cautious about adding another AI tool
02:19: Why Wyndo was initially pessimistic about Grok Bot
04:54: Multiple named Bots, cloud execution, skills, connections, and routines
08:37: Dheeraj challenges the laptop-off benefit and compares it with remote Claude Code and a private server
10:01: Sam asks about data privacy
11:47: How Wyndo gives each Bot a role, memories, connections, skills, and routines
15:10: Comparing Grok Bot with Manus and project-based assistants
17:44: The Bot as the main unit of work
19:16: Where scheduled cloud work fits
20:39: The first five Bots and their recurring jobs
23:58: Social Content Writer workflow through Typefully
25:09: Chief of Staff morning task preparation
25:51: Invoice Hunter workflow
26:21: Content Researcher workflow across YouTube, X, Notion, and Typefully
27:31: SEO Agent reporting and research workflow
28:57: Live tour of the Grok Bot interface
30:19: Finding and connecting apps through plugins
32:03: How skills are assigned to a Bot
33:37: Creating a new marketing Bot from a plain-language request
34:33: The simple progress view and limited visibility into the Bot’s reasoning
35:22: Editing the Social Manager’s role and profile
36:24: Sakita asks how the X API connection works
38:27: Authorizing X and Notion connections
41:07: Live Content Researcher test using Supadata and YouTube metrics
41:59: Social Manager demo through Typefully and Substack Notes
44:11: Business Coach and SEO Agent examples
45:10: Opening an August analytics report as HTML slides
47:12: Local computer access and shareable Bot templates
49:27: Live LinkedIn draft and scheduling request
50:51: Dheeraj describes Grok Bot as a guided agent-building experience
53:03: Model choice and the fit for less technical users
54:41: Current access and pricing questions
56:07: Dheeraj’s cost comparison with his existing n8n server
57:09: Where Wyndo currently uses Grok Bot, Claude Code, and ChatGPT
Resources Mentioned
Grok Bot: The cloud-based agent app used throughout the session. Its Bots can work through a browser, files, a command line, connected tools, skills, and routines. The same Bots are available on desktop and iOS.
Cursor pricing: Grok Bot access was listed with Cursor’s $20 per month Individual plan when checked on September 4, 2026. Higher plans offer higher usage limits, and taxes or usage charges may vary.
Grok Bot for iOS: The phone app Wyndo uses to message Bots, review work, and respond to approval requests away from his desk.
Grok Bot cloud computer and apps: The current documentation for the shared cloud computer, browser sessions, files, command-line credentials, and optional local-computer access.
Grok Bot skills and routines: Skills describe repeatable processes. Routines assign a process to a Bot and run it on a schedule or supported event.
Grok Bot approvals and privacy: The controls for sensitive inputs, consequential actions, and local-computer access.
Grok Bot templates: Shareable Bot configurations discussed near the end of the session. Wyndo also referenced a GitHub collection he used to discover and import Bot ideas. The exact repository link was not stated in the transcript.
Cursor: The account and subscription used to sign in to Grok Bot. Cursor’s coding product was also part of the session’s product comparison.
Claude Code: Wyndo and Dheeraj’s existing tool for local projects and deeper hands-on work. Dheeraj also discussed running Claude Code on his own server and using remote control.
Claude Cowork: Mentioned as a comparison for a guided project that can use uploaded or local files.
ChatGPT and ChatGPT Work: Compared with Grok Bot’s named roles, projects, cloud tasks, and computer-based workflows.
Codex: Mentioned alongside Claude Code and ChatGPT as another agent Wyndo uses for deeper project work.
Grok models: The model family behind the version of Grok Bot used during the stream. The interface did not offer Wyndo a model picker during the demo.
Grok Imagine: The image-generation feature used by Wyndo’s early Creative Director Bot.
DeepSeek: Mentioned as an open-model option covered in an earlier episode.
Gemini: One of the AI services Dheeraj already pays for and considered in his cost comparison.
Manus: Compared with Grok Bot because it offered cloud projects and mobile access, including a Telegram Bot.
Telegram: Mentioned as the mobile interface previously used with Manus.
Typefully: The scheduling service connected to the Social Content Writer. Wyndo used it for LinkedIn, X, and Substack Notes. Typefully’s July 27, 2026 changelog announced support for publishing and scheduling Substack Notes.
Notion: Wyndo’s content calendar and backlog, where the Content Researcher can place findings for review.
Supadata: The YouTube data and transcript service used by the Content Researcher during the live search.
YouTube: The source searched for Grok Bot videos and public engagement metrics.
X Developer Platform: The app and authorization layer Wyndo used to connect X. He discussed funding API access, but readers should check the current pricing before copying that setup.
X: Used as a research source, bookmark source, and publishing destination.
OpenSEO: The SEO research connection used for keyword and competitor data. OpenSEO estimates remain separate from first-party site performance.
Google Analytics: The first-party traffic source behind Wyndo’s weekly report and August 2026 HTML slides.
Google Search Console: The first-party search-performance source connected to the SEO Agent.
Todoist: The task app where the Chief of Staff can prepare Wyndo’s morning list.
Granola: Visible among the available Grok Bot connections during the interface tour. It was not used in the live workflows.
n8n: Dheeraj’s existing automation tool, hosted on his own server and used as the cost and control comparison near the end.
Whispr Flow: The voice-to-text tool Wyndo uses to speak instructions into Grok Bot on his phone.
MCP: The connection standard discussed for linking tools such as Notion, Typefully, Todoist, Supadata, and OpenSEO to an agent.
AI News Briefing, AI Use Digest, Weekly Google Analytics, and Write Social Post: Custom skills visible in Wyndo’s plugin list. Their setup and pricing were not discussed.
Marketing Head template: A shared Bot template Wyndo imported during the session to demonstrate how a configuration can be added to another account.
Claude Artifacts: Dheeraj used Artifacts as the comparison when he asked whether Wyndo could share the generated HTML report through a public link.
HTML slides: The file format the SEO Agent used to visualize Wyndo’s August analytics. The file could be opened in Grok Bot and downloaded for local use.
