AI can now take a description and turn it into a website or application in minutes. But building it is only part of the journey: something still needs to run it and make it available to users.
Modern hosting platforms can abstract much of that infrastructure away, so customers don’t need to manage the underlying technical details themselves. AI makes creation easier; the hosting layer makes the result live and manageable.
For hosting providers, the opportunity is to make the journey from “I want to build this” to “it’s live and working” as straightforward as possible.
That’s the role cPanel is addressing with cPanel AI, introduced as part of cPanel & WHM version 138. It brings AI-assisted creation, application deployment, troubleshooting and agent connectivity into the hosting environment.
For providers, the important question is what that means for the service they deliver to customers.
The new handoff between creation and hosting
AI changes where a hosting customer starts.
Instead of arriving with a finished HTML site or a traditional development workflow, a customer might arrive with an AI-generated site, a Git repository, or a Node.js application that was created from a prompt.
It may look finished. But there is still a gap between having the project and having a production-ready website or application.
The hosting platform has to bridge that gap.
That means handling things such as domains, DNS, SSL certificates, application runtimes, databases, resource limits and deployment. Just as importantly, it needs to tell the customer what is happening when something doesn’t go as planned.
For providers, that means supporting several paths to the same destination. A customer might build a site through an AI-assisted workflow, deploy an application they created somewhere else, or work directly from a Git repository. The experience can be different, but the result should be the same: a project that is live, manageable and understandable.
cPanel AI approaches this through Meridian, a goal-based interface organized around the things customers actually want to accomplish—websites, email, files, databases, security and performance.
Meridian dashboard — goal-based navigation across websites, email, files, databases, security and performance.
Meridian is available as an opt-in theme alongside the existing Jupiter interface. That means providers can introduce the new experience without altering live site data, DNS, or underlying configurations, and users can move between the two interfaces without downtime. For hosting providers with custom Jupiter plugins or proprietary integrations, existing workflows remain fully supported while Meridian’s native extension and plugin ecosystem rolls out in upcoming updates.
From an idea to a live site
For customers who don’t think in terms of runtimes, repositories and configuration files, the starting point needs to be simple.
Meridian’s Guided Setup brings several AI builder options together.
One of those options is Nova, an agentic AI app and website builder that is part of cPanel AI. A customer describes what they want, Nova builds it and shows a preview, and the customer publishes it to their hosting account when ready. Nova for WordPress follows the same approach but produces a standard WordPress site built with the Block Editor, which the customer can keep managing in WordPress Admin with any plugin they need. For providers, this means a customer’s first step into AI creation can happen inside the hosting account rather than on a third-party builder.
Guided Setup, step one: choose a temporary cpanel.site domain or bring an existing domain.
The customer can choose how they want to start rather than having to understand which technical components need to be configured first.
Guided Setup, step two: Nova (“Describe it, and Nova builds it”) sits alongside Nova for WordPress, WordPress, Sitejet, AI App Hosting and Custom Code.
The setup then brings the relevant choices together before launch.
Guided Setup, step three: review the domain and builder choice, then launch.
This matters because many customers who use AI to create a website aren’t going to think about environment variables, runtime versions or deployment configuration. They shouldn’t necessarily have to. The hosting platform can make those decisions automatically where it safely can, while still showing enough information for the customer—or a support team—to understand what happened when something needs attention.
Hosting has to accommodate applications, not just websites
The distinction between a website and a web application is becoming less obvious to customers.
A booking system, customer portal or dashboard can now start with a prompt just as easily as a marketing site can. The customer may not see much difference between the two. The hosting platform does.
Applications have runtime requirements, dependencies, build processes and processes that need to keep running.
That’s where AI App Hosting comes in. In cPanel & WHM v138, it provides a deployment path for Node.js applications. It can detect supported frameworks such as Next.js, Express.js, Nuxt and NestJS, automate supported dependency installation and build steps, run the application in an isolated Podman container, and route traffic through an Apache reverse proxy with automated HTTPS. The containers use upstream Node.js images rather than separate images for each framework, with the application source mounted into the container and the start command supplied at runtime.
Customers can deploy from a ZIP file or a Git repository, while the platform automates the supported build, runtime and routing steps. These deployments can be initiated through the Meridian UI or through MCP, giving customers and AI clients different ways to work with the same hosting capabilities.
Deploying a Node.js app: pointing AI App Hosting at a Git repository.
It’s important for providers to be clear about the scope. In v138, AI App Hosting is focused on pure Node.js deployments across both primary domains and subdomains. Existing Node.js applications deployed through Jupiter that use Passenger are based on a different backend technology and are not manageable through Meridian. Python, Ruby and Go workloads continue to use existing manual or standard deployment methods, with additional runtime options, deeper resource controls, and automated Git push workflows planned for future iterations.
The app live: certificate issued, process running.
That distinction is important for setting customer expectations. AI-assisted hosting doesn’t mean every application can be deployed without configuration. It means the hosting platform can take responsibility for more of the common deployment path and make the remaining requirements easier to understand.
Put help where the problem happens
Deployment is only part of the story.
Once a site or application is live, customers still have questions. Why isn’t the certificate working? What is using all the disk space? Why did an application stop? Is the DNS configuration correct?
Traditionally, the answer has been a knowledge base article, a support ticket or a conversation with a support team.
Those options aren’t going away, but AI can change where the first interaction happens.
In cPanel’s annual user survey, the most-requested capability was an assistant that could read logs, identify a problem and recommend what to do next.
That’s the idea behind the AI Assistant in Meridian. With the appropriate account context and permissions, it can use configuration information and logs to explain settings, guide customers through tasks and help diagnose issues involving areas such as SSL, DNS, email, disk usage, databases and certificates.
AI Assistant answering an account-specific question against the account’s actual domain data.
The important part isn’t simply that there is a chatbot in the control panel. It’s that assistance is available alongside the thing the customer is trying to manage.
For providers, that could reduce some of the repetitive work that reaches support teams. But the goal shouldn’t be to make the number of support conversations as small as possible. A customer who gets a confusing answer and gives up hasn’t had a successful support experience.
More useful measures include successful resolution, customer effort, successful deployments and, where appropriate, support-ticket deflection.
The boundaries matter too.
The AI Assistant is primarily conversational and diagnostic: it reads and explains what is on the customer’s account. When a task does call for a change, the Assistant proposes it and the customer approves it before anything runs. The change executes through cPanel’s own API, within the customer’s own account permissions, and it is recorded in an audit trail. Taking action is enabled separately from the Assistant itself, so it can be switched off while answering and diagnosing remain available.
When AI agents need access, permissions become part of the product
There is a significant difference between an AI assistant answering a question and an AI agent taking an action.
Once an agent can inspect accounts, modify configuration or initiate deployments, permissions become a core part of the hosting service.
Model Context Protocol, or MCP, provides a standard way for AI systems to connect to tools and data. But the protocol itself doesn’t remove the need for access controls. Providers still need to decide which servers, accounts, tools and actions an agent can reach.
The basic principle is familiar from other areas of infrastructure: start with the minimum access required and expand it deliberately.
Read-only access may be enough for diagnostics and reporting. Actions that can change infrastructure should be more narrowly scoped, authenticated and auditable.
cPanel’s MCP support reflects two different operational boundaries.
Customer-level MCP is scoped strictly to an individual customer’s account and home directory. Access is authenticated directly within Meridian, allowing an end user’s AI client to interact with their own account without requiring separate third-party credentials or granting server-wide administrative access.
Admin-level or fleet MCP is a separate path for server administrators. It is operated through the WebPros Dashboard, with its own permissions and visibility across a server fleet.
Keeping these boundaries separate is important. A customer asking an AI agent to inspect their website should not need the same access as an administrator managing hundreds of hosting accounts.
The broader principle is straightforward: AI should make infrastructure easier to manage without turning convenience into unrestricted administrative access.
Claude Code (CLI) over MCP, running a health check against a production cPanel server.
The same MCP session from Claude’s desktop app, generating a remediation checklist.
What hosting providers should consider
Before rolling out AI-assisted hosting, providers should answer a few practical questions.
- Which customers are you trying to help? A customer who wants to generate a simple website has very different needs from a developer deploying a Node.js application or an administrator managing a server fleet. Those workflows don’t necessarily need to be introduced at the same time.
- How tightly can access be controlled? Providers should be able to define access at the appropriate level—server, account, tool and action—and understand exactly what an AI system is permitted to do.
- Can you see what happened? When an AI agent takes an action, the audit trail should make it clear what happened and when. That becomes increasingly important as agents move beyond answering questions and begin interacting with live infrastructure.
- How will you measure whether it works? Successful deployments, successful resolutions, customer effort and support outcomes are more useful measures than simply counting how many times customers interacted with an AI feature.
A phased rollout gives providers a chance to answer these questions with real customer behavior rather than assumptions.
The hosting layer becomes the bridge
AI is making it dramatically easier to create things for the web. But creating something isn’t the same as running it.
Customers still need a production environment that connects domains, applications, email, security and ongoing management.
That puts the hosting layer in an interesting position.
The hosting provider can become the bridge between an AI-generated idea and a reliable production service.
Guided onboarding can help a customer move from a description to a working website. AI App Hosting can provide a deployment path for AI-generated Node.js applications. The AI Assistant can help customers understand and troubleshoot what is running. MCP can give AI agents a structured way to interact with hosting infrastructure while keeping provider-defined permissions in place.
None of this means that AI removes the need for hosting expertise. In many ways, it makes the hosting layer more important. As creation becomes easier, customers will expect the infrastructure underneath it to feel just as straightforward.
The most useful AI hosting experience won’t necessarily be the one with the most AI features.
It will be the one that makes the journey from idea to live, manageable service feel natural—while making it clear who has access to what, what the platform is doing, and when a human needs to step in.
