How to Use Cloudways to Scale a Server for a Campaign Launch

The short answer: if your site runs on a 4GB DigitalOcean Premium server at $54 a month, holding an 8GB server through a 72-hour launch adds 72 x ($0.1473 - $0.0804) = $4.8168 to the invoice, because Cloudways bills Flexible servers by the hour (pricing page, checked on 26 October 2026). The steps: confirm the server can come back down, let a read-only agent pull the last launch's load, rehearse on staging, put the CDN in front, scale up the day before, watch errors and bandwidth during the campaign, and scale back the morning after.

Disclosure: I may earn a commission through some links on this site. It does not change what this page says.

Work these out on paper first:

1. What the launch costs. The 72 hours above cost $4.8168. Keeping the 8GB server for the whole month instead costs $99 - $54 = $45 more, on the same pricing page. The hours are cheap; the forgotten weeks are not.
2. What the last launch did to the server. Your last campaign's busiest hour is the baseline. Cloudways keeps monitoring graphs per server, and its MCP server can read CPU, memory and traffic for a past day or week.
3. What "it held" means. Pick two readings before launch day: no 5xx errors on the landing page, and the app's slow-pages list looking like an ordinary day. Copilot flags 5xx errors caused by plugins, themes or the database, per its product page.

Vertical scaling makes the server you have larger rather than adding servers, and you pay for the larger size only while you hold it. This guide is about choosing those hours. Plans and limits are in the Cloudways glossary entry.

What you need

• A Cloudways Flexible server running the site. Autonomous plans add capacity on their own and do not need this guide.
• Access to the DNS records for your domain, for the CDN step.
• The hour the campaign goes live and the hour it ends.

Step 1: Confirm the way back down before going up

The Flexible hosting page describes plans you can scale up or down. The provider notes on the pricing page are narrower: they say AWS and Google Compute Engine support both upgrading and downgrading, and say nothing of the kind for DigitalOcean. If your server is on DigitalOcean, ask support in chat whether it can return to its current size after the launch. If the answer is no, the fallback is cloning to a smaller server afterwards, which MCP 2.0 lists as the server_clone tool. Settle this first, because it decides whether three days at 8GB stay a three-day cost.

Step 2: Let an agent read the last spike, read-only

This is the agent step. The Cloudways MCP server connects clients such as Claude, Cursor or ChatGPT to your account at https://mcp.cloudways.com/mcp/, with your account email in the X-CW-Email header, per the use-case post. Since MCP 2.0, every connection uses an access token generated under API Integration, and API keys stopped working on 15 October 2026. Generate a Read-Only token for this job and give it an expiry date.

Then ask for the evidence in Cloudways' own phrasing: "Show CPU and memory for server 12345 over the last 7 days", "Show slow PHP pages for app 67890 in the last hour", "Show slow MySQL queries for app 67890". Point it at the dates of your last campaign. Ask it to finish with a short sizing note: peak CPU and memory, the slowest pages, and the size it would choose.

The human check sits here. The agent reads and drafts; a person picks the size and makes the change in Step 5. A Read-Only token cannot scale, clone or delete anything, which is the point: MCP 2.0 places server create, scale and delete under Full Access.

Step 3: Rehearse the launch changes on staging

Every Flexible plan includes advanced staging and cloning, per the pricing page. Clone the application, make the launch-day changes there (the new landing page, any plugin it brings), and check the page, the form and the checkout before they reach production. A plugin that misbehaves on staging costs an afternoon; on launch morning it costs the campaign.

Step 4: Put Cloudflare Enterprise in front

The add-on serves whole pages from Cloudflare's edge, which takes load off the server before any scaling does. The add-on page sets out three moves: open Cloudflare in Application Management, add the domain and click Enable, then edit the CNAME or A records at your DNS provider. No separate Cloudflare account and no nameserver change are needed. It runs free for 30 days, then costs $4.99 per domain a month, and bandwidth past the included 100GB costs $0.02 per GB. Switch it on before launch week, so the DNS change and the traffic do not arrive on the same day.

Step 5: Scale up the day before

Open Server Management, then Vertical Scaling, and choose the larger size, as the FAQ on Cloudways' DigitalOcean hosting page describes. Cloudways' traffic spikes post says the size changes without moving to a new server. The billing follows at once: the pricing FAQ says a plan change shows on the invoice immediately, charged by the hour. Doing it a day early leaves time to click through the site at its new size while traffic is still ordinary.

Step 6: Watch two numbers during the campaign

Errors on the landing page, and bandwidth. Copilot alerts by email and in the platform when it detects an issue, per its product page. Bandwidth past the allowance costs $0.02 per GB on DigitalOcean and $0.12 on AWS, per the pricing page FAQ; the 8GB DigitalOcean server includes 5TB a month, so check its graph halfway through rather than on the invoice.

Step 7: Scale back and read the invoice

When the campaign ends, return to the old size, or clone to a smaller server if Step 1 ruled out going down. Ask the agent for the Step 2 readings again, so the next launch starts from this one's numbers. Invoices go out in the first week of the next month for the month before; check that the 8GB hours stop where the campaign did.

Warnings in Cloudways' own guidance

• Deletions are permanent. The use-case post says deleted servers and applications cannot be recovered, and that restoring an application is itself destructive. Confirm the target before any such command.
• Tokens reach further than they look. MCP 2.0 advises a Read-Only token for every new integration and notes that some read calls return application credentials, so even that token needs care.
• Old keys are dead. A client still set up with an API key lost access on 15 October 2026 until a token replaces it.
• Regions change the price. On AWS and Google Cloud the price depends on the region, per the pricing FAQ.

The free part comes first

Steps 1 to 4 cost nothing beyond the server you already pay for: a support question, a read-only agent, a staging clone and a CDN on a 30-day trial. They also answer the question most launch plans skip, which is whether the server needs to grow at all. If the edge cache carries the landing page and the Step 2 note shows headroom, keep the 4GB server at $54 and skip Step 5. Hourly billing cuts both ways, and Cloudways is paid on the way you forget: 72 hours at 8GB cost $4.8168, while the same server left at 8GB to the end of the month bills the full $45 difference on the pricing page rates, on an invoice that arrives weeks after the campaign.

FAQ

Can Cloudways add capacity automatically during a spike?

That is what Autonomous is for: managed WordPress on Kubernetes, from $99 a month on Growth, with capacity above the baseline charged at $0.07 per server-hour, per the pricing page. Flexible, the product in this guide, runs at the size you choose.

Does changing the server size take the site offline?

Cloudways' traffic spikes post says you can change the size without migrating or worrying about downtime. Do it a day early anyway.

Which token scope does the agent need?

Read-Only, for everything in this guide. Scaling through the agent would need Full Access, per MCP 2.0, which is why the scale-up here is a person's click in the dashboard.