At Faaaster, we have always had an obsession: that our customers never had to wait for an answer.
WordPress hosting high performance, this is our job. And when an e-commerce site falls or a page takes 6 seconds to load, every minute counts.
So we asked ourselves a simple question: What if our support AI could access our internal tools directly to diagnose a problem in real time?
Today, it's done. You are told how – and especially why – we built this technological brick.
The problem: a powerful, but restrained support
Before this development, our support was already working well. Our team knows WordPress and WooCommerce on the fingertips, and our Crisp chatbot was able to handle the most common requests.
But this is the classic scenario: a customer writes "My website has been slow since this morning". To respond, our team had to follow several manual steps. Log in to the admin dashboard, launch a PageSpeed report, check the server metrics (CPU, RAM, bandwidth), check the recent logs, and sometimes cross with the site configuration ( PHP version, installed extensions, disk space). The result: even for a simple diagnosis, it was necessary to juggle between several interfaces and assemble the pieces of the puzzle by hand. Multiplyed by dozens of tickets a day, it represented a considerable amount of time – and above all, a time the customer was waiting for.
We said we could do better.
The solution: an MCP that connects the AI to our systems
What is the MCP?
The MCP – for Model Context Protocol – is an open standard that allows an AI to interact in a structured and secure way with external tools. In concrete terms, it is a bridge between the artificial intelligence model and your business systems.
Think of it as an API, but designed specifically for AI agents. Instead of simply exposing endpoints, the MCP describes the available tools, their parameters, and the conditions under which the AI can use them. This allows the AI to "reason" on which tool to use, when, and how to interpret the results.
What we built
We developed a custom MCP server that displays a series of tools connected to our internal systems. Our Crisp chat L
Here's what our AI can question today:
- Performance diagnostic : L的IA can run a complete PageSpeed audit and retrieve the server metrics of the client's site: mobile and desktop scores, TTFB, LCP, CLS, but also CPU, RAM, and monitoring incident history.
- Billing and subscriptions : Check for unpaid invoices, invoices due, details of any invoice, and identification of risk subscriptions. No more need to ask the customer for his invoice number and pick him up manually.
- Logs and events Access to site logs on a specific time window, and complete event history: updates, configuration changes, incidents. When a customer says "it planted last night," the AI can check what happened exactly at that time.
- Technical configuration : Complete view of the site environment: extensions installed and their versions, version of CMS and PHP, available disk space, and much more. Ideal to identify an obsolete extension or version conflict.
Security at the Heart: The Human in the Loop
We're not going to lie: giving an AI access to customer data is not done lightly. This is why an approach has been put in place Human in the Loop rigorous.
How does it work in practice? When the AI needs access to a tool to meet a request, it cannot do so alone. The customer must be connected to his Faaaster dashboard and explicitly validate each tool use by the MCP. He's in control. The AI does not search the accounts independently: it asks, the client approves, and only then the diagnosis is started.
This secure validation guarantees two things. First, confidentiality: no access to data without the active consent of the site owner. Next, transparency: the customer sees exactly what tools are being requested and what information is being consulted.
What it really changes for our customers
Instant and complete answers
Before, a performance diagnosis could take several minutes of exchanges and require the intervention of a team member. Now, the client asks his question, validates access, and the IA gives him a structured diagnosis in seconds. Scores PageSpeed, server metric, log analysis, everything is synthesized in a clear and actionable response, directly in the chat.
24-hour support available
IA doesn't sleep, does not take a lunch break, and does not handle 4 conversations at the same time hoping to remember nothing. For our customers working on different time zones or facing a problem at night, this is a major change: a first level of diagnosis is always available.
A support team focused on essentials
Recurrent requests: "Is my website online?", "What is my bill?", "what version of PHP j Our team of experts can focus on what really requires human expertise: complex performance issues, migration, custom configurations, strategic advice.
The result is not a dehumanized support. It is an increased support, where humans intervene where they bring the most value.
Technical backstage: integration with Crisp
A word about our technological partner in this adventure: Crisp. Their customer support platform already offered a quality conversational AI, but it was their MCP custom integration that made all this possible.
Integration is articulated in several layers. The MCP server that we developed interfaces with our internal APIs – monitoring, billing, site management – and exposes these capabilities as standardized tools. On the Crisp side, conversational AI can call these tools contextually, depending on the client's demand. The MCP protocol ensures that exchanges are structured, typed, and secure.
The work of the Crisp team on this subject is welcomed. Their responsiveness, the quality of their technical support, and the robustness of their MCP integration were decisive. When you innovate, having a partner who moves at the same pace changes everything.
Now what? The continuation of the programme
What we have put in place today is only a first step. IA can consult, diagnose, and inform. But we already see the rest.
Our next step: exploit the capabilities of WordPress 7.0 to allow the AI to communicate directly with the customer's WordPress. Imagine: IA is no longer content to say "your extension X is obsolete", it can update it for you. It no longer only tells you "your cache is misconfigured", it adjusts the settings.
We move from a medium that informs to a medium that acts, always under the control of the client, of course.
Why it matters to the WordPress ecosystem
WordPress hosting is a market where many actors look alike. Same technical stacks, same performance promises, same dashboards. What makes the difference is the customer experience. And the customer experience, in 2026, goes through intelligent, fast, and transparent support.
At Faaaster, it is believed that the AI is not there to replace the customer relationship, it is there to enrich it. Give each customer the means to get an expert diagnosis in seconds, without technical barrier, without waiting time, this is our vision of modern WordPress hosting.
And we're just starting.
You use Faaaster and you want to test IA support? He's already active in your Crisp cat. Ask your question, and see the difference. And if you are not yet a customer, it may be time to discover what a really smart WordPress hosting can do for you → faaster.io
