Grok + xCloud

Manage servers with Grok on xCloud

Managing servers with Grok Build means typing short requests into xAI's terminal agent so that it reads your xCloud servers, adjusts firewall and Fail2Ban settings after your approval, and reports what actually changed.

Skill: xcloud:servers
Toolsets: servers, alerts, user
Free with every xCloud account

Set up Grok All Grok guides

YouAllow tcp 5432 on Frankfurt from 203.0.113.10 only.

servers_firewallRulesread-only

Agentfrankfurt-1 has four rules and none covers 5432. I would add: name postgres-office, tcp, allow, port 5432, source 203.0.113.10. Create it?

YouYes.

servers_firewallRules_createneeds approval

AgentThe task finished with no errors and the rule list now shows five rules.

Last updated: October 2026. Keep a human in the loop: xCloud stops and asks before anything that creates, deploys, updates, reboots, deletes or buys.

Setup

How Do You Set Up Grok to Manage servers on xCloud?

Connect Grok once; every job on this account uses the same connection. Then ask in plain words.

1
Add the xCloud MCP server
Run this in your terminal, then start grok. The first time it calls xCloud, a browser opens on the xCloud approval screen, where you tick the teams and choose Read-only or Full access. Add --scope project to save the entry in .grok/config.toml for one repository instead.
Terminal
```
grok mcp add --transport http xcloud https://app.xcloud.host/mcp
```
2
Or edit ~/.grok/config.toml
Add this table to ~/.grok/config.toml, then run grok mcp doctor xcloud to check the connection. The url key is the one for HTTP servers.
TOML
```
[mcp_servers.xcloud]
url = "https://app.xcloud.host/mcp"
```
3
No browser? Use an API key
For a headless machine, create a token with the mcp:invoke scope plus the read abilities for the areas it will use (read:servers and read:sites, with read:billing and read:addons for billing and add-on tools) and the matching write: abilities if it should change things in Settings, Developers, API Tokens. Put it in an environment variable the machine starts Grok with, such as XCLOUD_TOKEN in a systemd EnvironmentFile or your shell profile, and pass a reference to it with --header. Keep the single quotes so the shell leaves ${XCLOUD_TOKEN} alone: Grok expands ${VAR} in headers when it loads the config, so the token itself never lands in your shell history or in config.toml. Run grok mcp doctor xcloud to confirm it works.
Terminal
```
grok mcp add --transport http --header 'Authorization: Bearer ${XCLOUD_TOKEN}' xcloud https://app.xcloud.host/mcp
```
✓
Check it worked
Then ask Grok for the job itself, for example:
Prompt
```
Show Fail2Ban bans on Frankfurt. Unban 203.0.113.10.
```

In practice

How Does Server management Work from Grok?

Server security is where a terminal agent earns its place, because the request is often urgent and small. A teammate cannot reach the server over SSH, or a deploy script is being blocked, and you want an answer in the shell where you already are. In Grok Build you type: show Fail2Ban bans on Frankfurt. The agent calls servers_fail2ban_banned-ips, prints the addresses and, if one of them is the person you are chasing, you say unban it. The agent names the server and the address, asks, and only then calls servers_fail2ban_unban. If the SSH restriction is the real cause, servers_firewall_ssh-restriction-status shows how SSH access is restricted before anyone edits a rule.

Firewall edits take a few more words and a bit more care. Grok Build lists the current rules with servers_firewallRules, then proposes the new one in full: a name, the protocol, allow or deny, the port and an optional source address. Nothing is created until you answer. Disabling a rule is easier to undo than deleting it, so ask for a disable when you only want to test a change. xCloud can also whitelist an address for you. Ask Grok Build to whitelist your current IP, and it calls servers_firewall_whitelist-caller-ip. The caller address is the one xCloud sees on that request, so read the address the agent reports before you approve, and use xCloud's own IPs option when you mean the platform and not yourself.

Connection work shows up in the same terminal. grok mcp doctor xcloud checks the connection, and /mcps opens the MCP tab inside a session, which helps when the question is whether Grok Build has the server tools at all. Sudo users follow the same approval pattern, with one difference you should know about: a password is a secret and private keys are never returned, so the agent will not echo either back to you.

Grok specific: Register xCloud in ~/.grok/config.toml, not with --scope project. Servers do not belong to a repository, and a project-scoped entry in .grok/config.toml makes the xCloud tools disappear when you start grok in another directory. Firewall and SSH changes can lock you out of the very machine you are typing on, so ask Grok Build to show the rule and read the SSH restriction status before it creates or enables anything, and whitelist your own address first when you are about to tighten access.

What xCloud does for server management

xCloud lets the agent list your servers, read their monitoring, and manage services, Node.js and PHP versions, cron jobs, firewall rules, Fail2Ban and sudo users. Changes that can interrupt a server stop for your approval, and a reboot is only reported as done after xCloud verifies a new boot.

Pick the team and the server. The agent works on one team at a time and uses the server you name. If a name is missing or matches several servers, it lists them and asks. It always restates which server it is about to change before it changes it.
Read the state first. It reads the server's status, installed services, runtimes and monitoring figures. A status such as low disk space or reboot required is worth surfacing on its own. Monitoring history is a paid-plan feature, so a free plan answers 403.
Say what will change, then ask. For a service, runtime, cron, firewall or sudo change, the agent names the server, the item and the impact, for example that a Node.js default affects every Node site on the server. It waits for your approval when the change could interrupt a service.
Follow the task to the end. xCloud accepts server work and returns before it finishes. The agent reads the server's task list until the task settles and reports the real outcome, not the acceptance.
Reboot with proof. A reboot starts a tracked operation. The agent reads that operation until xCloud reports a verified new boot. If the result is unconfirmed it investigates and never repeats the reboot just to check.
Buy a server only after approval. A new server is billable. The agent checks what you already have, reads the plans and the card on file, shows the plan, region and price, and waits for a yes. It sends the purchase once with an idempotency key so a retry cannot buy a second server.

Reference

Server management Settings and Limits on xCloud

The facts Grok works within when it manages servers. Where a row names the dashboard, that step stays yours to take there.

Setting or limit	What applies
Inventory	Every server on the team with its status, stack and the sites it hosts, plus the tasks, supervisor processes and site snapshots on a server
Monitoring	Current CPU, memory and disk are readable. Monitoring history is a paid-plan feature: xCloud answers 403 on the free plan, which is a plan limit and not a permission
Services	Install, enable, restart and disable. A restart runs without a server-side prompt; the agent still confirms the exact server, service and impact before any service change, because disabling ssh, nginx or a database can lock you out or cause downtime
Node.js versions	Read the installed versions and set the default. The default is server-wide and affects every Node site on the server
PHP versions	Install, uninstall, set the default, patch and toggle OPcache. The server default changes only the command-line php and the version new sites get; it does not change existing sites, and a single site's PHP version is a dashboard setting under Site > Site Settings
Cron jobs	List, create, update, delete, run now and read the last output for server cron jobs. Site cron is a separate resource
Firewall rules	List, create, delete, enable and disable rules with a name, protocol, allow or deny, port and optional source IP
Fail2Ban and IP access	List, ban and unban IP addresses, read the SSH restriction status, and whitelist your current IP or xCloud's own IPs
Sudo users	List, create or update and delete sudo users. A password is a secret and private keys are never returned
Verified reboots	Start a tracked reboot and read its operation; only a verified new boot proves it worked. An unconfirmed reboot can be rechecked without rebooting again
Buying a server	Billable, so it needs an approved plan, region and price, a card on file and an idempotency key. It buys xCloud-managed servers only
Dashboard-only	Resizing or deleting a server, a provider backup of the whole server (Server > Backup), bringing your own server, databases and database users, and PHP settings such as memory limit or upload size

Rules Grok has to follow

A change that can interrupt a server stops for your approval, and the agent names the server, the service and the impact first.
Buying a server is billable: the agent shows the plan, region and price and waits for your yes, and it never retries a purchase without checking your server list first.
A reboot is only reported as done when xCloud verifies a new boot, and an unconfirmed reboot is investigated, not repeated.
Disabling ssh, nginx, a database or a runtime service needs your explicit confirmation immediately before the call, because it can lock you out or take sites offline.
Resizing or deleting a server, provider server backups, bringing your own server and per-site PHP versions are dashboard steps; the agent gives you the path and the dashboard link xCloud returned.

Example prompts

What Can You Ask Grok to Do for Server management?

Type these as written and swap in your own repository, site and server names. Reads and routine actions such as backups, cache purges, PageSpeed scans and vulnerability scans run straight away; creating, deploying, updating, rebooting, deleting, buying or starting a broken-link scan stops and asks first.

Prompt

Show Fail2Ban bans on Frankfurt. Unban 203.0.113.10.

Prompt

List Frankfurt firewall rules. Add allow tcp 5432 from 203.0.113.10 and show it to me before you create it.

Prompt

Whitelist my current IP on Frankfurt and tell me which address you whitelisted.

Prompt

List all my xCloud servers with CPU, RAM and disk usage, and flag any server above 80% disk.

Prompt

List the cron jobs on the Frankfurt server with their schedules.

Prompt

Create a server cron job for the WooCommerce Action Scheduler every five minutes. Show me the final command and schedule before creating it.

Prompt

Install Redis on the Frankfurt server and enable the service.

Prompt

Switch the Frankfurt server's default Node.js to the latest LTS major.

Prompt

Reboot the staging server and tell me when it is back.

Prompt

Show me the IP addresses Fail2Ban has banned on the Frankfurt server, and unban 203.0.113.10.

Prompt

Create a new xCloud server on the smallest plan in Singapore. Show me the plan and price first and wait for my approval.

Grok and Server management: Frequently Asked Questions

What people ask before they let Grok manage servers through xCloud.

Can Grok Build unban an IP address that Fail2Ban blocked?

Yes. Grok Build reads the banned addresses on the server, names the address it will unban and asks before it calls the unban tool. You can also ask it to whitelist your current IP, and you should check the address it reports before you approve.

Can a firewall change from Grok Build lock me out of my own server?

It can if a rule blocks your own connection, which is why the agent shows the full rule and reads the SSH restriction status first. xCloud lets you disable a rule instead of deleting it, which is easier to reverse. Disabling ssh itself needs your explicit confirmation immediately before the call.

Does the agent ask before it reboots a server?

Yes. A reboot stops for your explicit confirmation, then the agent starts a tracked reboot and reads the operation until xCloud reports a verified new boot. If the result stays unconfirmed it investigates rather than rebooting a second time.

Can an AI agent buy a new server for me?

Yes, but only after you approve it. A new server is billable, so the agent first checks your existing servers, reads the available plans and your card on file, then shows the plan, region and price and waits for your yes. It sends the purchase once, so a retry cannot buy a second server.

If the agent changes the server's PHP default, do my sites change version?

No. The server default changes the command-line php and the version new sites get, and it moves no existing site. Each site keeps its own PHP version, which you change in the dashboard under Site > Site Settings.

Can the agent resize or delete a server?

No. Resizing and deleting a server are dashboard-only, as are a provider backup of the whole server, databases and bringing your own server. The agent can tell you where the control is and give you the server's dashboard link.

Why did the agent say monitoring history is not available?

Monitoring history is a paid-plan feature, so xCloud answers 403 on the free plan. Current CPU, memory and disk readings are still available, and the agent will say that it is a plan limit, not a missing permission.

Other agents