Claude Cowork + xCloud

Secure sites and servers with Claude Cowork on xCloud

Securing sites and servers with Claude Cowork means delegating the whole review to Anthropic's desktop agent, which works through your xCloud sites and servers step by step and comes back with a report and any changes waiting for your approval.

Skill: xcloud:servers
Toolsets: vulnerabilities, servers, sites, wordpress-actions
Free with every xCloud account

Set up Claude Cowork All Claude Cowork guides

YouReview security across my team. Rescan the worst sites and report back.

vulnerabilities_indexread-only

servers_sudoUsers_indexread-only

AgentEleven sites checked. shop.example.com still shows one critical finding after the rescan. The Frankfurt server has a sudo user that looks temporary. I have not changed anything.

YouPrepare the plugin updates for the shop.

AgentThree updates are ready. I will take a backup first and wait for your approval before I update.

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 Claude Cowork to Secure sites and servers on xCloud?

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

1
Add xCloud as a custom connector
In Cowork, open Customize, then Connectors, and choose Add custom connector. If you already added xCloud to your Claude account, look for it in that list first. Name it xcloud, use this URL and sign in when the xCloud approval page opens.
URL
```
https://app.xcloud.host/mcp
```
2
Check the connection
Start a task and ask this first. Claude Cowork calls xCloud and names your account and the teams you approved, so you know the connector works before you delegate real work.
Prompt
```
Who am I on xCloud?
```
3
Match the access level to the job
Optional. On the xCloud approval screen, Full access lets a delegated job deploy, update plugins and renew certificates; Read-only is enough for audits and reports. To change the level later, reconnect the connector and approve again. A Team or Enterprise owner adds the same URL under Organization settings, then Connectors.
URL
```
https://app.xcloud.host/mcp
```

Check it worked

Then ask Claude Cowork for the job itself, for example:

Prompt

Run a security review of every site and server on my xCloud team. Rescan the worst three sites, compare the counts, and give me a short report with fixes waiting for my approval.

In practice

How Does Security Work from Claude Cowork?

A security review is a good fit for delegation because it is long, repetitive and mostly reading. In Cowork you do not trade one question for one answer. You describe the outcome, for example a security review of every site and server on the team, hand it over and go back to your own work. Cowork starts with the team-wide vulnerability rollup, picks the sites with the worst findings, starts a rescan on each and then waits, because a rescan finishes asynchronously and a count read too soon is out of date. It then reads the counts again and compares them with the first pass.

The second half of the task is the perimeter. For each server Cowork reads the firewall rules, the Fail2Ban ban list, the SSH restriction state and the sudo users. It flags what looks odd, such as a rule that opens a database port to the whole internet or a sudo user that looks temporary and unused, and it keeps those notes for the report. What you get back is a single summary you can read in a couple of minutes: what is open, what changed since the rescan, which findings look like false positives and which fixes are plugin or theme updates. Nothing changes while you are away unless you approved Full access, and even then xCloud stops for confirmation on updates, rebooting and similar steps.

That gap between the task and the decision is the point of Cowork here. It can propose the remediation, such as updating three plugins after a backup, and leave it waiting for you to say yes. It can propose banning an address or disabling a rule and name the server and the effect, and still wait. Where a protection is dashboard-only, such as the AI bot blocker or the 8G firewall, the report says so and gives you the dashboard path. Claude Cowork reaches xCloud through the custom connector from Anthropic's cloud, so the review does not need your laptop open to your servers.

Claude Cowork specific: Cowork works on its own while you are away, so choose the access level before you delegate. Read-only is enough to read findings, firewall rules, bans and sudo users, but anything that starts a scan or changes a rule is a write that needs Full access. If a task stalls on a refusal, reconnect the connector with the level you intend and approve again, then ask Cowork to pick the review up where it left off.

What xCloud does for security

xCloud reads vulnerability findings for one site or across your whole team, starts rescans and ignores or restores a finding. On servers it manages firewall rules, Fail2Ban bans, the SSH restriction state and sudo users. Protections such as the AI bot blocker, the 7G and 8G firewall and per-site IP rules stay dashboard switches.

Start with the team-wide rollup. For a fleet question the agent reads the vulnerability rollup across every site on the team in one call and sorts it worst first. For one site it reads the findings and the counts by severity instead.
Rescan and triage findings. A rescan runs without a confirmation stop and finishes asynchronously, so the agent checks the count again afterwards. Findings carry a severity and a source, and severity is the triage key. A false positive can be ignored with a reason and restored later.
Fix what the finding points to. Most findings are fixed by updating a plugin, a theme or WordPress core. The agent proposes the updates, takes a backup first if you want one, and waits for your approval, because updating plugins or themes always stops for confirmation.
Review the server's perimeter. The agent reads firewall rules, the Fail2Ban banned IP list, the SSH restriction status and the sudo users. It can add or disable a rule, ban or unban an address and whitelist xCloud's own IPs after you confirm the server and the effect.
Report what is dashboard-only. AI bot blocking, the 7G and 8G firewall and a single site's IP allow and block list are dashboard switches. The agent reads what it can, such as the 7G and 8G logs, and gives you the dashboard path and the site's dashboard link for the rest.

Reference

Security Settings and Limits on xCloud

The facts Claude Cowork works within when it secures sites and servers. Where a row names the dashboard, that step stays yours to take there.

Setting or limit	What applies
Vulnerability findings	Per-site list and counts by severity (critical, high, medium, low), plus a team-wide rollup across every site you can reach
Rescans	Triggering a rescan is a routine action that runs without a confirmation stop. It is asynchronous, so the agent polls the count or the site events
Ignore and restore	A finding can be ignored with a reason and un-ignored later. Both are changes to the finding, not to the site
Firewall rules	List, create, enable, disable and delete server rules by name, protocol (tcp or udp), traffic (allow or deny), port and optional source IP
Fail2Ban	List banned IPs, ban one or more addresses and unban an address. Adding addresses to Fail2Ban's own ignore list is an SSH step described in the Fail2Ban guide
Whitelisting	xCloud's infrastructure IPs and the caller's IP can be whitelisted. The SSH restriction status says whether SSH is limited to trusted addresses
Sudo users	List, create or update, and delete OS-level privileged accounts. Private keys are never returned, and a temporary account is available for short jobs
7G and 8G firewall	Switched on in the dashboard. Their logs are readable through the nginx log type of the site's access log read
AI bot blocker	A dashboard toggle: Site, Tools, Nginx and Security on Nginx, Tools, Security on OpenLiteSpeed
Per-site IP rules	Dashboard-only: Site, Tools, IP Management. The API lists a site's rules but does not change them; server-wide blocks are on the API
Automated security updates	xCloud installs security updates on servers it provisions, and Automatic Reboot restarts the server only when a reboot is required. Both are configured in the dashboard
Security headers	The API lists a site's custom Nginx configuration. Adding headers is a dashboard or SSH step, described in the security headers guide

Rules Claude Cowork has to follow

Firewall, Fail2Ban and sudo-user changes name the exact server and the effect before they run, and they wait for your yes. Locking out SSH is the risk the agent explains first.
The agent never disables SSH or a database service to tidy up a ruleset, and it does not delete the default SSH rule on its own.
Updating a plugin or theme to fix a finding stops for your approval, and the agent offers a backup first.
Ignoring a finding needs a reason you give. The agent records it and tells you how to restore it.
AI bot blocking, the 7G and 8G firewall, per-site IP rules and automated update settings are dashboard steps. The agent says so, gives the path and the dashboard link, and never claims it changed them.
A temporary sudo user is for a specific job and is removed when the job ends.

Example prompts

What Can You Ask Claude Cowork to Do for Security?

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

Run a security review of every site and server on my xCloud team. Rescan the worst three sites, compare the counts, and give me a short report with fixes waiting for my approval.

Prompt

Each Monday, check Fail2Ban bans and firewall rules on all my servers and tell me anything that changed or looks unusual.

Prompt

Go through the sudo users on every server, flag any that look temporary or unused, and do not remove anything until I approve it.

Prompt

Show the vulnerability count for every site on this team, sorted worst first, and name the top three findings.

Prompt

Run a vulnerability scan on the shop site and tell me what it finds.

Prompt

List the findings on the blog site and ignore the false positive in the contact form plugin, with the reason.

Prompt

Update every plugin on the shop site, then run a vulnerability scan, and stop if anything is still open.

Prompt

List the firewall rules on the Frankfurt server and tell me whether SSH is restricted to trusted IPs.

Prompt

Show the IPs Fail2Ban has banned on the Frankfurt server, and unban 203.0.113.10.

Prompt

Whitelist the xCloud IPs on the Frankfurt server.

Prompt

Which sudo users exist on the Frankfurt server? Flag any that look temporary or unused.

Claude Cowork and Security: Frequently Asked Questions

What people ask before they let Claude Cowork secure sites and servers through xCloud.

Can Claude Cowork run a security review on a schedule?

You can hand it the same review as a repeatable task, and it will read findings, rules and bans each time you start it. Anything it proposes still stops for your confirmation, so check the report before approving updates or rule changes.

Will Claude Cowork change my firewall while I am away?

Only if you approved Full access, and even then xCloud and the agent name the server and the effect and wait for your yes. With Read-only, Cowork can review rules and bans and report, but it cannot create, disable or delete a rule.

Can an agent show vulnerabilities across all my sites at once?

Yes. xCloud has a team-wide rollup that lists findings for every site you can reach, which the agent sorts worst first. For one site it reads the findings and the counts by severity, and it can start a rescan.

Can the agent ban an IP address on my server?

Yes. It can ban addresses in Fail2Ban, list the banned IPs and unban one, and it can create, disable and delete server firewall rules. It confirms the server and the effect first, so you do not lock yourself out.

Can the agent turn on the 8G firewall or the AI bot blocker?

No. Both are dashboard switches. The agent can read the 7G and 8G firewall logs through the site's nginx logs, and it tells you where the switches are: the site's Tools section. It does not claim to have changed them.

Can the agent block an IP for just one site?

Not from here. Per-site IP allow and block lists are managed in the dashboard under Site, Tools, IP Management, and the agent can only list what is there. Server-wide blocks, through firewall rules and Fail2Ban, are on the API.

Does the agent fix a vulnerability on its own?

No. It reports the finding and proposes the update that remediates it, usually a plugin, theme or core update. Updates always stop for your approval, and the agent can take a backup first and rescan afterwards to confirm the finding is gone.

Other agents