How to set up Grok Bot and chain specialist bots

This is the setup path: stand up the desktop app, name a specialist with one job, hand it a real task, then chain the rest on the shared cloud computer.

Check plan, OS, and privacy

You need an eligible plan—SuperGrok Plus, SuperGrok Heavy, Cursor Pro+, Cursor Ultra, or Cursor Teams Standard or Premium (sign in with your Cursor account)—the Grok Bot desktop app for macOS, Windows, or Linux, and an app or website where the first Bot can do useful work. You can also set up Grok Bot on mobile; the computers Bots work on run in Cursor's cloud.

Grok Bot requires cloud data storage. Accounts using Legacy Privacy Mode must move to a supported Cursor data setting before it can start. Review that at Cursor privacy settings.

Install the desktop app

Open the Grok Bot downloads page and choose the download for your computer.

macOS: Choose Apple silicon or Intel, open the disk image, drag Grok Bot to Applications, then open it. If macOS asks for confirmation, choose Open. Check Apple menu → About This Mac: a Chip field means Apple silicon; a Processor field means Intel.
Windows: Choose x64 or Arm64, run the installer, and open Grok Bot from the Start menu. Architecture is under Settings → System → About → System type.
Linux: Builds publish with every stable release under More downloads. Choose x64 or Arm64 as a .deb for Debian and Ubuntu, an .rpm for Fedora and RHEL, or an AppImage. Install with your package manager, or make the AppImage executable and run it, then open Grok Bot from your application launcher. Check architecture with uname -m: x86_64 is x64, aarch64 is Arm64.

Grok Bot checks for updates automatically. You can also use Check for Updates from Settings → Beta.

Choose Get started on the welcome screen, or Sign In with Cursor from Settings. Finish authentication in the browser, then return to Grok Bot. It uses your Cursor account; complete the normal organization SSO flow if required.

On first use it introduces Bots, the shared computer, and routines, then asks which tools you use. Those answers shape the first teammate suggestions; they do not connect or modify those tools by themselves. Computer setup runs in the background. The last step opens Meet a future teammate.

Create a named Bot with one job

Choose a suggested teammate, or Create your own with a short name, one primary job, and a description of how it should work. Focused Bots build more useful context than one catch-all. Add more later with New → Create new agent when work splits into distinct roles.

Get-started's example:

Name: Piper

Job: Product performance

Description: Investigate product-performance questions using our observability tools. Preserve links and screenshots, separate evidence from hypotheses, and return a short summary with the highest-impact issue first. Never change production settings.

Grok Bot 101 matches that anatomy as three fields—Name, Title, and Description—set in chat or in Settings, top right. Be as specific as you can. Describe the workflow the way you would do it, or record a demo on your screen.

Hand off a first task

A strong request includes:

Outcome: What should be finished?
Sources: Which apps, websites, files, or conversations matter?
Constraints: What must the Bot avoid or ask before doing?
Deliverable: What should it return?
Review point: When should it stop for you?

For a five-minute first result with no connector or website login, attach a document and try:

Summarize this document in five bullets. List every date, decision, and open question in a separate section. Cite the page or section for each item. Do not change the source file.

Then try work in one of your tools, still naming the outcome, sources, constraints, deliverable, and review point. Before allowing external changes, review Approvals, security, and privacy. Ask the Bot to revise anything incomplete, name lasting preferences explicitly, and when the process is stable, save it as a skill or create a routine.

Take over the Agent Computer

When the Bot hits an app that needs authentication, it may ask you to take over. Open Agent Computer from the conversation, choose the takeover control, enter the password, passkey, two-factor code, or complete the CAPTCHA yourself, then return control.

The browser session persists, so other Bots can use the same signed-in session when appropriate. For supported services, install a connector from Settings → Plugins and authenticate it in your browser. Use connectors where available and computer use for everything else.

All Bots share one cloud computer—files, browser sessions, and app logins. The computer belongs to your account, not to an individual Bot; treat anything on it as available to every Bot you run. Between users, isolation is strict. If you log into a site with one bot, every other bot can reach that site too. Write permission rules in natural language in Settings > General > Agent. A separate review agent can allow, block, or escalate proposed actions.

Chat, routines, and specialist chains

Three ways to talk to a bot:

Chat: Send it a message.
Routines and triggers: It can set a schedule or listen for events, including a Slack thread or a GitHub PR.
Other bots: Bots can message each other and trigger each other.

Treat them as specialists. Drop several into a group chat, or let a weekday routine route a request through different bots until the job is done. In Grok Bot 101, a Marketplace bot watches Facebook Marketplace and Craigslist; when it needs a judgment call, it asks a Chief of Staff bot.

Bots can run in parallel, share context in group chats, and pass ownership of a task. Each Bot gets its own screen on the shared computer, and one Bot runs one computer-use task on its screen at a time. Closing the app, laptop, or phone does not stop a background turn or a routine.

Lingxi Li’s engineer fleet: launch, inspect, retry

Lingxi Li documents a fleet of specialist engineer bots that manage Cursor cloud agents. Each bot specializes in a different area and performs best focused on a single domain. Every bot in that fleet can create Cursor cloud agents, read transcripts, review proofs attached to PRs, and send follow-ups by queueing a message or interrupting the run.

When they receive a task, they kick off a cloud agent with skills invoked and a thorough prompt covering the work and expected proof. Grok Bot can monitor the transcript and artifacts such as screenshots, notify you when the run is done, queue messages, or interrupt if anything goes wrong. If an agent hits environment flakiness and stalls, the bot stays on the run and unblocks it as aggressively as it can.

Install from x.ai/bot, create one named specialist with a single job, and send the five-part first task. After you take over the Agent Computer for a login, add a second specialist and let them message each other.