Grok Bot for coding works as a coding teammate that sits outside your editor. A Bot runs commands in a terminal on its cloud computer, reads your repositories through the GitHub connector, and starts routines when GitHub, Sentry or Linear events happen. It can also hand bigger code changes to Cursor Cloud Agents, which can open a pull request for you to review. This page shows which xAI-documented feature to use for each coding job, where you approve the work, and where Grok Bot is the wrong tool.
📺 Watch: Grok Bot + Claude is INSANE!
🔥 Get the Agent OS as a free bonus: AI Profit Boardroom members get the full Agent OS zip, prompt libraries, daily tutorials and weekly live coaching calls. → Get inside
Want ready-made AI agent workflows for your business, plus live calls to get them running? Join the AI Profit Boardroom. If you would rather plan your coding setup with us first, book a free AI strategy session.
How Grok Bot for coding is set up
There is no separate Grok Bot account. According to xAI's docs, you sign in with your Cursor account. If your access comes from a seat on a self-serve Grok Business plan, you choose Sign in with Grok instead. The computers Bots work on run in Cursor's cloud.
The key detail for coding: all of your Bots share one persistent cloud computer, with a browser, filesystem and terminal. Each Bot gets its own screen on that shared computer, so several Bots can work at once. One Bot can run one computer-use task on its screen at a time. Because the computer is shared, files, browser sessions and command line credentials are available to every Bot on your account.
Coding work reaches you in four ways:
- Terminal on the cloud computer. A Bot can run command line tools there, the way you would in your own terminal. A terminal command card shows Allow and Skip when Auto Review wants your decision.
- Cursor Cloud Agents. A Bot can delegate a coding task to a Cloud Agent, which runs on a separate computer under your team's Cloud Agent controls. When it finishes, the Bot attaches the cited files to its reply. View PR opens the pull request if the run opened one, and Open in Cursor opens the run in the Cursor app or on the web.
- Projects. The Grok Bot changelog (v0.57.0, 18 September 2026) says a Bot can start a Project, "a Cloud Agent that plans the work and runs its own agents".
- Your own computer. Running commands on the Mac or Windows machine in front of you is a separate setting, called Execution on this computer. The default is Ask every time.
Team admins control delegation. The Cloud Agents switch on the Grok Bot page is on by default, and an admin can turn it off for the whole team.
Grok Bot for coding: task, feature, trigger and approval point
| Coding task | Grok Bot feature | Trigger | Approval point |
|---|---|---|---|
| Catch up on a repo or a pull request | GitHub connector (list repos, compare commits, list a PR's commits, read branch and issue comments) | You ask in chat, or a scheduled routine | Read-only summary in chat. Nothing to approve |
| Triage a new production error | Routine on a Sentry or PagerDuty event | A new Sentry or PagerDuty event | An Ask first rule before the Bot posts or changes anything |
| Reproduce a reported bug | Routine on a Slack message, plus browser and test account on the cloud computer | A Slack message that matches your rule | "Never post back to Slack without approval" in the routine |
| Ship a code change | Delegation to a Cursor Cloud Agent | You ask, or a GitHub or Linear event routine | You review the pull request through View PR before merging |
| Run a script or a test suite | Terminal on the shared cloud computer | You ask in chat | Terminal command card: Allow or Skip |
| Run a command on your laptop | Execution on this computer | You ask in chat | Ask every time (default), per command |
📺 Watch: Grok Bot Is INSANE! 3X Faster AI Agents?!
Sample prompts for Grok Bot for coding
Each prompt names the feature it relies on. Copy one, then swap in your repo, channel or project names.
- GitHub connector: "Using the GitHub connector, compare the last ten commits on main in the website repo with last Friday's state. List what changed and flag anything that touches checkout."
- Pull request summary: "List the commits in the open pull request on the booking-app repo, read the issue comments, and give me a three-line summary plus open questions for the author."
- Sentry routine: "Create a routine: when a new Sentry event arrives for the api project, find the matching commit and draft a short incident note in this chat. Ask first before posting anywhere else."
- Bug repro routine: "When a message in #support contains a ticket link and the phrase needs repro, reproduce the issue in staging with the test account and post a repro pack in this conversation. Never post back to Slack without approval."
- Cloud Agent delegation: "Hand this to a Cloud Agent: fix the date format bug on the invoice page, add a test for it, and open a pull request. Attach the screenshots from its report."
- Save a skill: "Save the process we just used as a skill called Weekly dependency check. Include the repos, the output format and the rule that you never merge anything."
- Narrow auto-approval: "Add a custom rule: allow automatically when running git status in the reports folder. Ask first for every other command."
Before you rely on a new routine, use Test to run it once. A new routine waits for its next scheduled time, and only events that happen after the trigger was created will start it.
📺 Watch: Every Level of Grok Bot Explained
Who Grok Bot for coding fits, and who it doesn't
It fits
- Small teams that already use Cursor and want a teammate who watches GitHub, Sentry, Linear or Slack while they work.
- Founders and agency owners who want code changes to arrive as pull requests they can review from a phone.
- Teams that want glue work handled for them: repo summaries, bug repro, release notes and incident notes.
It doesn't fit
- Developers who want an AI pair programmer inside their editor. Grok Bot is a chat-and-cloud-computer teammate, not an editor extension. For in-editor work, compare Grok Bot vs Cursor and Grok Bot vs Copilot.
- Teams that must choose their model. Grok Bot has no model picker and picks the model for each task itself.
- Teams that need each agent fenced off from the others. All your Bots share one computer, and xAI's docs say not to use separate Bots as a security boundary.
- Companies that need on-premises hosting. Grok Bot runs only on Cursor-hosted cloud computers.
Grok Bot for coding limits to plan around
- Shared credentials. Command line credentials on the cloud computer are available to every Bot on your account. xAI advises signing out of services you no longer need and removing sensitive temporary files when the work is done.
- Auto Review is a model. xAI's docs say Auto Review should complement least privilege and explicit approval boundaries, not replace them. Keep rules narrow, like the git status example, rather than "allow everything".
- Separate event connections. Event triggers come from Cursor account integrations. These are separate from the Slack or GitHub apps and may need their own connection.
- GitHub connector scope. The v0.75.2 changelog entry (22 October 2026) lists reading actions: repos, commit comparisons, PR commits, and branch and issue comments. Plan to route code changes through Cloud Agents and pull requests.
- Delegation can be switched off. If your admin turns off Cloud Agents, Bots cannot hand coding work to them.
Grok Bot for coding compared with other coding agents
Grok Bot is closest to an async teammate. Dedicated coding agents sit nearer the code. For autonomous engineering agents, read Grok Bot vs Devin. For a terminal-first coding agent, read Grok Bot vs Claude Code. If you prefer an open-source agent you host yourself, Hermes Agent for coding covers that route. Our roundup of the best AI agents for coding ranks the wider field.
The model behind a coding agent matters a lot. Julian's own model tests are in Goldie Bench. To see how coding agents slot into a full business system, read the Agent OS guide.
Want help deciding where AI agents fit in your dev workflow? Book a free AI strategy session, or get the step-by-step AI agent workflows inside the AI Profit Boardroom.
Grok Bot for coding FAQ
Does Grok Bot plug into my code editor?
xAI's docs do not describe an editor plugin. You work with Bots in chat on desktop, mobile, Slack or X. Results from a Cloud Agent run can be opened in the Cursor app or on the web with Open in Cursor.
Does each Bot get its own computer?
No. All of your Bots share one persistent cloud computer, and each Bot gets its own screen on it. Delegated Cloud Agent work runs on separate computers under your Cloud Agent controls.
Can Grok Bot open pull requests?
Through Cursor Cloud Agents, yes. When a delegated run opens a pull request, View PR opens it from the Bot's reply, so you can review it before anything merges.
Can I pick which model writes the code?
No. Grok Bot has no model picker and chooses the model for each task. If model choice matters to you, book a free AI strategy session and we will help you compare AI agents that let you pick.
