Yes, your Grok Bots share one computer. xAI and Cursor's docs say each user gets one persistent cloud computer, and every named Bot on that account uses it, including its files, browser sessions and logins. Each Bot has its own screen, its own memory and its own conversations, but they aren't walled off from each other. xAI's FAQ is blunt about it: "Do not use separate Bots as a security boundary."

How does the shared computer work?

Cursor's Grok Bot docs describe a persistent cloud computer with a browser, a filesystem and a terminal, running in Cursor's cloud. The computer belongs to your account, not to any one Bot. xAI's teams documentation says it's a Firecracker microVM, isolated from other users.

Shared across your Bots Separate per Bot
Files Yes
Browser sessions and logins Yes
The computer itself Yes
Screen Yes, one each
Memory and preferences Yes
Conversations Yes
Computer-use tasks One at a time per Bot

Each Bot runs one computer-use task at a time on its own screen. Several Bots can work in parallel on the same machine.

Why does launch marketing say "their own computer"?

xAI's launch materials described Bots as having "their own computer". The docs say something different: one computer per account, shared by all your Bots. Our research found the two statements in conflict, and the docs are the more detailed and more recent source. Plenty of third-party coverage still repeats the launch line, so check the date on anything you read.

One way to read the launch phrase is that your Bots have their own computer as a group, separate from your laptop. That's true. It isn't true for each Bot on its own.

What do named Bots keep separately?

Each named Bot keeps its own memory, preferences, role context and summaries of past work, according to the Cursor docs. Its conversations stay separate too. That's what makes a "research Bot" and an "inbox Bot" feel like different teammates.

Context still travels between them in three ways: shared files, shared browser sessions, and handoffs, where one Bot passes ownership of a task to another. Bots can also message each other and share context in group chats of two to six Bots.

Cloud computer per user
1
Bots per group chat
2 to 6
Computer-use task at a time per Bot
1
Data removal after deleting Grok Bot
30 days

Why does sharing matter for security?

Because a Bot that reads malicious instructions on a web page or in an email can reach everything on the computer, not just its own work. Its files, browser sessions and logins are the same ones your other Bots use. Cybernews and Digital Applied both flag this shared-computer risk.

That's why safeguards work at the account level. Approval prompts ("Allow once", "Deny" and "Always allow" on desktop), Auto Review and takeover for passwords, 2FA, CAPTCHAs and payments apply across your Bots. The docs say Auto Review "should complement, not replace, least privilege and explicit approval boundaries." Our sister site's guide to least privilege for AI agents covers that approach. For background on the attack itself, see prompt injection explained.

Your own desktop is a separate question. Bots only run commands on your local machine if you allow it in Settings, then Computer, where the default is "Ask every time". See Grok Bot computer access settings.

Do team Bots use my computer?

No. xAI's teams documentation says shared and team Bots in Slack and group chats run on their own computer, not a member's. Team Bots can join Slack channels with their own handle and use plugins such as Salesforce, Notion and GitHub. That separation is useful: a team Bot shouldn't be able to reach a member's personal logins.

Why does each Bot ask me to sign in again?

If the computer and its logins are shared, you might expect to sign in once. Users on the Cursor forum report otherwise: each Bot's browser still needs a separate sign-in despite the account-level login promise. A thread titled "Confused over Grok Bot" from 19 October has 22 replies. xAI hasn't explained the mismatch in the docs we reviewed, so expect to repeat sign-ins through takeover until it does.

What happens when you delete a Bot?

Deleting one Bot is a soft delete, according to xAI's FAQ. Its files and logins on the shared computer may remain, because they belong to the computer rather than to that Bot. If you want them gone, remove them yourself or sign out of the sites involved.

Deleting Grok Bot entirely is a bigger step: it deletes your Cursor account, and data is removed within 30 days.

Pros

  • Bots can share files and handoffs without copying data around
  • Each Bot keeps its own memory, role and conversations
  • Team Bots run on their own computer, away from members' logins

Cons

  • Separate Bots are not a security boundary
  • Deleting one Bot can leave its files and logins behind
  • Users report repeated sign-ins per Bot despite shared logins

How does this compare with Meta Muse?

Meta Muse also runs each user's agent on a separate cloud VM, but it adds Sentinel, a separate agent on the same machine that is the sole authority for connector actions and network traffic leaving the VM. Grok Bot relies on approvals, Auto Review and takeover instead. Our Grok Bot vs Meta Muse comparison covers the trade-offs, and how personal AI agents work explains cloud computers more generally.

Bottom line

All your named Grok Bots share one cloud computer, with the same files, browser sessions and logins, while each keeps its own memory, screen and conversations. Treat them as one security unit, not several. Team Bots are the exception: they run on their own computer. For the rest of what Grok Bot does, see our Grok Bot review and the setup guide.