Windsurf + xCloud

Speed up a slow site with Windsurf on xCloud

Speeding up a slow site with Windsurf means asking the Devin Local agent in your editor why a site is slow, and letting xCloud answer from CPU, RAM and disk samples, cache settings, PageSpeed results and the access log.

Skill: xcloud:performance
Toolsets: pagespeed, sites, servers, broken-links
Free with every xCloud account

Set up Windsurf All Windsurf guides

YouWhy is shop.example.com slow? Use xCloud.

sites_statusread-only

sites_cacheSettingsread-only

sites_pagespeed_latestread-only

AgentThe site is not mid-deploy. All three cache layers are off and the mobile run shows a slow server response. Turn on page cache under Site, WordPress, Caching; the link is in the panel. I cannot enable it from here.

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 Windsurf to Speed up a slow site on xCloud?

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

1
Add xCloud to the Devin Local agent
New tabs in Devin Desktop use the Devin Local agent, which reads MCP servers from the Devin CLI config files. Run this in a terminal: the URL is treated as Streamable HTTP, and the second command opens the browser for the xCloud sign-in (the agent also prompts on first use). By default the entry lands in .devin/mcp_config.local.json for the current project; add -s user to the first command to share it across projects in ~/.config/devin/mcp_config.json, where the entry reads url plus transport http.
Shell
```
devin mcp add xcloud https://app.xcloud.host/mcp
devin mcp login xcloud
```
2
Or edit the legacy Cascade config
If your tab runs the legacy Cascade agent, click the three-dot menu in the Cascade panel, then the Open MCP config file icon in the MCPs section, and add this under mcpServers. Cascade allows 100 tools in total and the full xCloud server offers 188 operations plus two search tools, so point serverUrl at the compact profile, five tools that reach every operation through search and call; a single toolset such as ?toolsets=sites (60 tools) also fits, but sites and servers together are 121 tools. Windsurf's file has been at ~/.codeium/windsurf/mcp_config.json, and the current documentation lists ~/.config/devin/mcp_config.json on macOS and Linux and %APPDATA%\devin\mcp_config.json on Windows; the icon opens the one your version reads.
JSON
```
{
  "mcpServers": {
    "xcloud": {
      "serverUrl": "https://app.xcloud.host/mcp?profile=compact"
    }
  }
}
```
3
No browser sign-in? Use an API key
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, export it as XCLOUD_TOKEN and add a headers field to the xcloud entry. Both agents fill in the ${env:XCLOUD_TOKEN} reference from your environment, so the token itself stays out of the file. The Devin Local entry is shown; for Cascade the same headers field sits beside serverUrl.
JSON
```
"xcloud": {
  "url": "https://app.xcloud.host/mcp",
  "transport": "http",
  "headers": {
    "Authorization": "Bearer ${env:XCLOUD_TOKEN}"
  }
}
```

Check it worked

Then ask Windsurf for the job itself, for example:

Prompt

Why is shop.example.com slow? Use xCloud to read the cache settings, monitoring history and latest PageSpeed result.

In practice

How Does Performance Work from Windsurf?

You are in Windsurf with the site's code open, the agent chat docked beside it, and a slow page you cannot explain from the source alone. That is where this job helps. You type the question, why is shop.example.com slow, and add that the agent should use xCloud. It resolves the site and its server, checks that the site is not mid-deploy, reads the CPU, RAM and disk history, calls sites_cacheSettings, reads the latest PageSpeed run, looks at the access log, checks php-fpm, the web server and Redis, and reads the PHP version. Each call appears in the chat as a tool call, so you can see what was measured before you read the conclusion. These are reads, so xCloud adds no prompt of its own, though Devin Local asks before an MCP tool runs by default. If those prompts get in the way, its permissions allow list takes patterns such as mcp__xcloud__* or mcp__xcloud__sites_status.

The editor adds one thing a chat app does not: the code is right there. xCloud's data names the likely cause, such as an uncached site, a nearly full disk or one client hitting a path, and then you can decide what to do about it in the project you have open. The agent does not guess at application bugs from the measurements. If the numbers are all normal it says so, and it does not pretend the cause is in your theme. When the fix is a dashboard switch, like page cache under Site, WordPress, Caching, the agent names the path and gives the dashboard link, because the API reads cache layers and purges them but cannot enable one.

The connection shapes this job on only one agent. Connect the full URL, https://app.xcloud.host/mcp, with devin mcp add xcloud and devin mcp login xcloud, and Devin Local has what it needs, since no tool cap is documented for it. If your tab runs the legacy Cascade agent, it can use at most 100 tools across all servers, and the full xCloud server offers 188 operations plus two search tools, so connect the compact profile. Narrowing Cascade by toolset does not rescue this job: the diagnosis reads from both sites and servers, which count 60 and 61 operations, so the two together already pass the cap before PageSpeed is added.

Windsurf specific: New Windsurf tabs use the Devin Local agent, and Cognition documents the Cascade MCP page as applying to legacy Cascade only, so if xCloud tools do not appear in a performance chat, check which agent your tab runs first. Devin Local reads ~/.config/devin/mcp_config.json, .devin/mcp_config.json or .devin/mcp_config.local.json, and on versions before 3.6 it reads the main config.json and migrates entries on its own. A Cascade tab opens its file from Open MCP config file in the Cascade panel menu and is held to 100 tools. For Cascade the cause is usually the tool cap, and for either agent it can be the wrong config file, not the xCloud connection.

What xCloud does for performance

xCloud reads the site's and the server's CPU, RAM and disk samples, which cache layers are on, the latest PageSpeed result, traffic in the access log, service health and the site's PHP version. The agent names the cause from that data, then hands you the switches that only the dashboard can flip.

Resolve the site and its server. The agent looks up the site and the server it runs on, because the answer often comes from both. A slow site is frequently slow because a neighbour on the same server is.
Check status and monitoring history. A site that is mid-deploy or failed explains slowness without any measurement. Otherwise the agent reads the site's and the server's samples, and a 24-hour or 7-day series to tell whether the slowdown is new. Disk comes early: a nearly full disk slows everything, the database first.
Read the cache layers. The agent reads the cache settings before saying anything about caching: whether page cache, Redis or Object Cache Pro object cache and Cloudflare edge cache are on, and which server stack the site runs. It never reports that no cache configuration is readable without calling it.
Read PageSpeed and traffic. The latest completed PageSpeed run is free to read. A new scan is worth starting only when there is no result or the site changed since, and the agent tells you it is spending a scan. The access log shows spikes, a crawl or one client hammering a path.
Check services and PHP. The agent checks that PHP, the web server, Redis and the database are running, and reads the site's PHP version and pending updates. A stopped Redis next to an object cache that reads as on is a found answer.
Name the cause and hand off. You get the most likely cause with its numbers. The fix is either something the agent can do, such as a purge after you agree, or a dashboard switch it names with the path and the site's dashboard link. If every number looks normal, it says so instead of inventing a cause.

Reference

Performance Settings and Limits on xCloud

The facts Windsurf works within when it speeds up a slow site. Where a row names the dashboard, that step stays yours to take there.

Setting or limit	What applies
Monitoring history	Site and server CPU, RAM and disk samples as a 24-hour or 7-day series. History is not available on the free plan, and the agent says that is a plan limit rather than reporting no data
Cache state	Read through sites.cacheSettings: page cache, object cache (Redis, Object Cache Pro) and Cloudflare edge cache. It returns settings only, never cache contents
Enabling a cache layer	Dashboard-only: Site, WordPress, Caching. The API reads the layers and purges them, and cannot switch one on for an existing site
Page cache details	Cache duration, URL and cookie exclusions and ignored query parameters are dashboard-only and are not returned by the cache settings read
PageSpeed	The latest mobile and desktop results and the history are read for free. One scan runs both strategies
Scan conflicts	A 409 while a scan is running means it is still pending or scanning. The agent polls it instead of starting another, and it is not a cooldown
Cache purge	Purging runs without a confirmation stop but throws away a warm cache, so the agent asks first on a live site and reports each layer as queued or skipped
Traffic	The access log shows spikes and bots. It is read over SSH, so the agent asks for a bounded window
PHP version	Read from the site's status. Changing one site's PHP version is a dashboard step: Site, Site Settings
Server PHP default	Changing the server default PHP changes only the command-line php and the version new sites get. It does not move any existing site
Broken links	A broken-link scan on a WordPress site returns findings grouped by source page, and a truncated scan is reported as partial

Rules Windsurf has to follow

The agent diagnoses from measurements, not guesses, and it calls the cache settings read before it says anything about caching.
It never offers to turn on Redis or any other cache layer itself. Enabling a layer is a dashboard step, and the agent gives you the path and the dashboard link.
It never presents a change of server PHP default as a substitute for changing one site's PHP version.
A cache purge on a live site, a new PageSpeed scan and any server-level PHP change are announced first. Server-level changes also need your explicit yes.
A site that errors rather than loads slowly goes to troubleshooting, and a failed deploy goes to deploy.

Example prompts

What Can You Ask Windsurf to Do for Performance?

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

Why is shop.example.com slow? Use xCloud to read the cache settings, monitoring history and latest PageSpeed result.

Prompt

Which PHP version does the blog site run, and are there pending updates that could be part of the slowdown?

Prompt

Read the last 200 nginx access log lines for shop.example.com and tell me if one client or path stands out.

Prompt

Why is shop.example.com slow? Check the cache, the server load and the latest PageSpeed result.

Prompt

Is Redis object cache on for the shop site? If it is off, tell me where to switch it on.

Prompt

Show CPU, RAM and disk for the shop site over the last 24 hours and tell me whether the slowdown is new.

Prompt

Run a PageSpeed scan on the shop site and compare the score with the previous scans.

Prompt

Look at the access log of the shop site for a traffic spike or a bot hammering one path.

Prompt

Which PHP version is the blog site running, and are there pending updates I should apply?

Prompt

Purge all caches on the shop site.

Prompt

Scan the shop site for broken links and list the pages with the most.

Windsurf and Performance: Frequently Asked Questions

What people ask before they let Windsurf speed up a slow site through xCloud.

Why would Windsurf need the compact profile for a slow-site check?

Only the legacy Cascade agent does. A slow-site check reads from both the sites and servers toolsets, which have 60 and 61 operations, and Cascade can use at most 100 tools across all servers. Point its serverUrl at https://app.xcloud.host/mcp?profile=compact, which exposes five tools, and it still reaches the same operations. Devin Local has no documented tool cap and can take the full URL.

Can Windsurf's agent fix my slow site from the editor?

It can find the likely cause from xCloud's measurements and purge caches after you agree. It cannot switch on a cache layer or change one site's PHP version, which are dashboard steps, so it names the path and links the site's dashboard.

Can the agent turn on caching for my slow site?

No. It can read which cache layers are on and purge them, but switching page cache, Redis object cache or Cloudflare edge cache on for an existing site is a dashboard step under Site, WordPress, Caching. The agent tells you the path and gives you the site's dashboard link.

What does a 409 mean when the agent starts a PageSpeed scan?

It means a scan for that site is still pending or running. The agent polls the existing scan instead of starting another, and it is not a cooldown period. A new scan is only worth starting when there is no recent result or the site has changed.

Will changing the server's PHP default speed up my site?

No. The server default changes only the command-line php and the version new sites get, and it does not move any existing site. The PHP version of one site is changed in the dashboard under Site, Site Settings.

Why does the agent say monitoring history is unavailable?

Monitoring history is a paid-plan feature, so on the free plan the request is refused as a plan limit. The agent reports it that way rather than saying there is no data, and it falls back to the samples and measurements the plan does include.

Does purging the cache ask for my confirmation?

xCloud itself runs a purge without a confirmation stop, because it is a routine action. The agent still asks you first on a live site, since a purge throws away a warm cache and the next visitors wait while it rebuilds.

Other agents