What Is Grok Bot? SpaceXAI's Always-On AI Teammates, Explained

§ 01What is Grok Bot?

In SpaceXAI’s own definition, a Bot is “a single persistent, named agent or one AI teammate.” You create a Bot by giving it a short name, one primary job, and a description of how it should work. Then you message it the way you would message a colleague, hand it a task with context, and it goes off to do the work.

The product started as an internal prototype. SpaceXAI says teams inside the company created Bots to handle sales outbound, marketing campaigns, office operations, and bug fixes before the public beta opened. That origin shows in the launch framing: the pitch is not a better chat window, it is delegation. In the company’s words, Bots “finish jobs end to end, and only come back when something needs your approval.”

Grok Bot is a separate application from the Grok chatbot. You install a dedicated desktop app (or the iOS app), sign in with a Cursor account, and manage team-level controls from the Cursor dashboard. The underlying model family is Grok, with Grok 4.6 released one day after the Bot launch, positioned for coding, agentic tasks, and knowledge work.

§ 02How the shared cloud computer works

The core architectural idea is that Bots do not run inside your laptop. Each user gets a managed Linux virtual machine in the cloud, with a browser, a filesystem, and a terminal. Bots use connectors and MCP servers where clean integrations exist, and computer use (driving real interfaces) for apps and websites that have no API. Work lands in the actual tool, not as a chat draft.

The detail that matters most: all of your Bots share that one computer. Files, browser sessions, and app logins are pooled at the account level, which is what makes handoffs between Bots cheap. Each Bot gets its own screen on the shared machine, so several Bots can run browser and desktop tasks in parallel. SpaceXAI is unusually direct about the consequence, stating in its documentation: “Do not use separate Bots as a security boundary.”

Bots can also act on your local computer: run commands, read files, and move files between the cloud machine and your own. The first local action asks for consent, and every local action then goes through an approval card that shows the exact command.

§ 03Skills, routines, and memory

Two mechanisms make Bot work repeatable. A skill describes how to perform a task. A routine is a saved path that can re-run on a schedule or on demand. The notable feature is learning by demonstration: ask a Bot to follow along once while you walk through a multi-step, multi-system process, and it persists that path as a routine.

On memory, a Bot retains stable preferences, role context, and summaries of prior work. Its conversation and learned role are separate from other Bots, while shared files and browser sessions can move context between them. The docs add a caveat worth reading twice: “For important decisions, ask the Bot to check the current source rather than relying on memory.”

§ 04What it costs and who can get it

As of August 14, 2026, there is no standalone Grok Bot subscription. The published access routes are:

Subscriptions include weekly usage, and eligible accounts can add on-demand usage billed from model and token cost. Team and enterprise rollout is in progress: enterprise organizations join through a waitlist via their Cursor account team.

Platform support at launch: macOS (Apple silicon and Intel), Windows (x64 and Arm64), and iPhone on iOS 18 or later. There is no Linux desktop app, no Android, and no iPad. Grok Bot also requires cloud data storage, so accounts on Cursor’s Legacy Privacy Mode must change that setting before Bots can start.

§ 05What Grok Bot does well

Read generously, and the launch materials earn it, Grok Bot gets several hard things right.

• It finishes work in real tools. The difference between a draft in a chat window and an updated CRM record is the whole product. Computer use plus connectors means the work lands where a human would put it.
• Zero-setup delegation. There is no workflow builder. You create a Bot and message it, and the same Bot is reachable from desktop and phone.
• Parallel Bots with a manager pattern. SpaceXAI describes people running a chief-of-staff Bot on top of specialists for inbox, expenses, recruiting, and operations. Bots message each other and pass ownership so you are not the router.
• Demonstration learning. Teaching a routine by doing the task once, with the Bot following along, is a genuinely low-friction way to encode a process.

§ 06The honest caveats

The same documentation is candid about the limits, and they matter for anyone evaluating the beta.

• One computer per user, not per Bot. Logins, sessions, and files are shared across all your Bots. SpaceXAI’s own words: separate Bots are not a security boundary. Anything one Bot can reach, every Bot can reach.
• Memory is advisory. The docs recommend re-checking current sources for important decisions rather than trusting what a Bot remembers.
• It is an early beta. Access is gated to three premium plans, enterprise is waitlisted, and coverage of the first days describes rough edges as expected in a controlled rollout.
• Platform gaps. No Linux, no Android, no iPad, and a hard requirement on cloud data storage.

None of these are unusual for a v1. They are worth knowing because the product’s promise (hand real logins and real work to autonomous software) is exactly the kind of promise where the security model and the memory model are the product.

§ 07Where Grok Bot sits in the AI employee category

Grok Bot’s use-case gallery reads like a job board: Sales Outbound, Talent Scout, Paid Media, Expense Manager, Account Health. That is the AI employee frame, arriving from the largest possible direction, and it validates the category more loudly than any startup could.

The differences are real, though. An AI employee, as we define it in our five-part test, owns a standing role with continuity, initiative, agency, and accountability. Grok Bot delivers strong agency and initiative. The gaps are role accountability (there is no task board, KPI, or standing report an owner reviews), memory that the vendor itself advises double-checking, and per-worker isolation. For a deeper cut on that contrast, see Grok Bot vs AI employees.

For contrast, CellCog has been building AI employees on Cellular Multi-Agents since October 2025: each employee runs in its own isolated workspace with its own email inbox, task board, shifts, and structured memory, so what it learns in one shift carries into the next, and it reports to you like a hire. Plans start at $8 a month, and a full-time employee typically runs $500 to 5,000 a month, usage-based at roughly $25 per shift. Different architecture, same destination: software you can give real work to.