Grok Bot is xAI’s AI agent product: AI teammates that you message like colleagues and that work on a computer of their own in the cloud. xAI announced it on 11 August 2026 as a beta, and its pages now describe apps for desktop and phone and a place in the paid plans of The X Subscription. If you searched for Grokbot, this is the same product.
This guide covers what the product is, how it works, who can use it, what it costs and where it falls short. Every vendor fact comes from the maker’s own pages, read on 22 October 2026, and plans and prices can change. DigitaVision is not affiliated with xAI, Cursor or Grok.
Key takeaways
- Each agent works on a computer in the cloud with a browser, files and a terminal, so work continues when your laptop is closed.
- The desktop app runs on macOS, Windows and Linux, and the mobile app runs on iPhone, iPad and Android, per the documentation.
- On the pricing page, the agent first appears on the Premium plan at $30 a month, and Super and Ultra add more usage.
- The platform chooses the model that serves each request, and the app has no model picker.
- The agents hand logins, two-factor codes and payments back to you, and sensitive actions can wait for your approval.
What is Grok Bot?
The launch post opens with a one-line pitch: “Grok Bot is your team of always-on agents.” xAI says they “have their own computer, work inside tools and apps like you do, and keep working 24/7.” You create a Bot, give it a name, a job and a short description of how it should work, then message it the way you would message a colleague. It finishes jobs end to end in the actual tool and comes back when something needs your approval, so the result is work done in the tool, not a chat draft.
The product began as an internal prototype. According to xAI, teams at the company built agents for sales outbound, marketing campaigns, office operations and bug fixes before the product was opened to everyone, and the documentation was last updated in October 2026.
How Grok Bot works: one hosted computer, parallel agents and routines
A shared computer of its own
The documentation says the product “works from a persistent cloud computer” with a browser, a command line, files and connected tools. An agent uses connectors where one exists and computer use for everything else, so it can sign in to sites that have no API. Everything on your account shares that computer, which means browser sessions, files and command-line credentials are visible to each of them. Each one gets its own screen, so several can work in parallel, although one runs a single computer-use task at a time. The Teams and Enterprises page adds that each member gets one dedicated Firecracker micro virtual machine, hosted in the US.
Teammates that run in parallel
You can run several agents at once, such as a researcher, a writer and a chief of staff, and put them in a group chat where they pass work and assign ownership. xAI says they can message each other and share context, so you are not the middleman between tools.
Routines: teach a workflow once
Ask an agent to follow along the next time you do a job. It watches the steps, saves the workflow as a routine and runs it again on its own, taking your corrections. The FAQ describes a skill as how to do a task, and a routine as a skill assigned to one agent with a schedule. The recording feature is limited to ten minutes, and its rollout may be gradual.
Memory, X and example jobs
A named agent keeps preferences, role context and summaries of prior work. For important decisions, the documentation suggests asking it to check the current source instead of trusting memory. Example jobs on xAI’s pages include a prospecting agent that researches accounts and drafts email and LinkedIn messages in your voice, a sales agent that updates the CRM with call notes and drafts follow-ups, an inbox manager for crowded inboxes, a customer-support agent and an office manager. You can also start work from X by tagging @bot, which needs a connected X account and is not available in every region yet.
Who can use it, and on which devices
The documentation lists the desktop app for macOS on Apple silicon and Intel, Windows on x64 and Arm64, and Linux on x64 and Arm64 as a .deb package, an .rpm package or an AppImage. The mobile app needs iOS 18 or later or Android 9 or later, and the iPhone app also runs on iPad. The launch post named desktop and iOS only, so the list has grown since August.
You sign in with a Cursor account. Access comes with every paid individual plan there and with the Teams plan, or by linking an individual SuperGrok Plus or SuperGrok Heavy subscription. The 26 August post lists Pro, Pro+, Ultra, Teams Standard and Teams Premium among the included plans. The X Subscription announcement of 12 October lists the agent as part of the new plan set. The pages differ a little on enterprises: the launch post offered a waitlist, the x.ai page says it is generally available for Enterprise, and the FAQ says enterprise access is rolling out. Ask xAI if you need it for a company.
Grok Bot pricing in October 2026 and what each plan includes
xAI sells the product inside The X Subscription, whose plans below come from the pricing page, before tax and subject to change.
| Plan | Monthly | Usage | Agent access |
|---|---|---|---|
| Free | $0 | Limited | Not listed |
| Plus | $8 | Shared pool | Not listed |
| Premium | $30 | 4x Plus | Listed |
| Super | $100 | 15x Plus | Listed, with priority model access at peak times |
| Ultra | $200 | 50x Plus | Listed |
The same page says each tier has a pool of AI usage shared across both products, and that the pool covers Chat, Imagine, Voice and Bot. Your allowance refreshes weekly, you can keep going with top-ups, and settings show what is left. The August posts said the agents have usage of their own, separate from your plans, and the FAQ says eligible accounts can add on-demand usage billed from model and token cost. Because the pages describe the pool in different ways, trust the usage screen in your account. No page we read states a separate price for the agents alone, and none lists a free trial; our free trial guide covers what xAI’s free plan includes. For the older plan names and pool rules, read our plans guide.
Which model does it use?
xAI’s documentation says Cursor manages model selection, that there is no customer-facing model picker, and that the serving mix can change over time. Usage analytics show which model served each request. So we cannot name a single model here, and any blog that does is not quoting these pages.
Limits and risks to know before you start
- One computer for all: every agent on your account can reach the same files, logins and sessions, so xAI says not to use separate agents as a security boundary.
- Websites can say no: a site may block automation, expire a session or ask for a human step, and the agent hands that step to you.
- Approvals expire: an approval request from work that started without you, such as a routine, expires after about ten minutes and the action does not run.
- Prompt injection: web pages and command output can try to steer an agent. The security page says its defences reduce but do not eliminate that risk.
- Privacy setting: the product needs cloud data storage, so Legacy Privacy Mode is not supported.
- Not every side effect is reviewed: Auto Review does not check everything, and memory writes are one example.
Your first handoff: how to get started with Grok Bot
- Pick an eligible plan: a paid individual plan on your sign-in account, or a linked subscription.
- Install the desktop app from x.ai/bot, or the mobile app from the App Store or Google Play.
- Sign in in the browser. If your access comes from a subscription, link it when the app asks.
- Choose a suggested agent or create your own with a short name, one primary job and a description.
- Give it a first task with an outcome, the sources, the limits and a review point, and take over the computer yourself for passwords and codes.
The overview page suggests a first task that spans a few tools and has a clear finish line, with drafts left for your approval. The approvals article recommends starting with read-only tasks and keeping sending, publishing, purchasing and deleting behind approval. To compare other assistants first, see our alternatives guide.
Grok Bot compared with Claude Cowork and ChatGPT Work
The closest rivals are Anthropic’s Claude Cowork and OpenAI’s ChatGPT Work. The table uses each maker’s own help pages and describes how the products are set up, not which is better at your job. Anthropic says Cowork uses the same agentic architecture that powers Claude Code, and xAI’s announcement bundles its agent with coding agents.
| xAI agent | Claude Cowork | ChatGPT Work | |
|---|---|---|---|
| Where work runs | One persistent hosted computer per user, used by all your agents | An isolated cloud environment created for each session | Cloud on web and mobile; on desktop, local files and apps with your permission |
| Platforms | macOS, Windows, Linux, iPhone, iPad, Android | Claude Desktop for macOS and Windows, web, mobile | Web, mobile and the desktop app |
| Who gets it | Paid individual plans or a linked subscription | Pro, Max, Team and Enterprise | Eligible paid plans |
| Model choice | Chosen by the platform, no picker | Not covered in the pages we read | Model picker in the app |
| Hand-off style | Message a named agent, routines and shared threads | Describe an outcome, projects and scheduled tasks | Describe the outcome, scheduled tasks |
The main design difference is persistence. An agent keeps its own computer and logins between tasks, while Anthropic describes a Cowork session as a temporary environment that is removed when the session ends. OpenAI’s help page adds that the older ChatGPT agent is no longer available and points to ChatGPT Work. Only xAI presents them as named teammates that coordinate with each other.
Is Grok Bot right for you, and who it is not for
It suits people who hand off recurring multi-step jobs across web apps, such as account research, email triage, expense handling or bug reproduction, and who already hold an eligible subscription. It is a weaker fit if you want a free tool, a single flat price for the agents alone, or a promise about which model answers. It also asks for care if you cannot share one hosted computer between everything you automate. If you are only curious, start with a small read-only task and watch what it asks you to approve.
The agents are bought from xAI or Cursor, and DigitaVision does not sell them. Its shared-seat listing, at $3.49 per 30 days, is a shared seat for conversations, so read the tool page for what it includes before relying on it for agent work, and browse other AI tools in our catalog.
FAQ
What is Grok Bot?
It is xAI’s agent product: named agents that you message and that work on a persistent computer in the cloud with a browser, files and a terminal. It was announced as a beta on 11 August 2026.
Is Grok Bot free?
The pricing page does not list it on the Free or Plus plans. It appears on Premium at $30 a month and above, and xAI’s pages list no separate trial.
Does Grok Bot work on iPhone and Android?
Yes, per the documentation: iPhone, iPad and Android, with the minimum versions listed above. The desktop app covers macOS, Windows and Linux.
Can Grok Bot keep working when my laptop is closed?
Yes. The FAQ says the work runs in the cloud, so closing the app, laptop or phone does not stop a background task or routine.
Which model does Grok Bot use?
xAI’s pages say the platform manages model selection and there is no model picker, so the serving model can change over time.

