Your AI assistant doesn’t know what’s going on in your organization today. Not because the model is flawed, but because it doesn’t have access to your systems. Model Context Protocol (MCP) changes that. Cube builds your MCP server: the connection between your AI agent and the systems that power your organization. This gives your assistant context, and you get accurate answers.
.jpg&output=webp&q=90&hash=MTc3NjI2MDgwOA==&w=800&h=800&fit=cover&a=center)
You can see what AI can do. But between the demo and daily practice, there's a gap.
Ask about the status of a project or an outstanding invoice. The answer sounds plausible, but it's wrong. Not because AI falls short, but because it has no visibility into what's actually in your systems right now.
Colleagues start supplying context themselves, copying data into external tools, or dropping out after the pilot. Not with bad intentions, but without a secure connection to your own systems, that's the only route left.
Controlled access to the data your AI agent needs. Your systems, your rules, your infrastructure. Cube builds that connection. From integrating existing systems to setting up the right access layers: we make sure your AI agent works with live business data, not assumptions.
The MCP server is at the heart of every integration between AI and your systems. Cube builds that server on top of your existing software. Your business logic remains intact. The MCP server makes that logic accessible to AI models without requiring you to overhaul your architecture. Your existing systems continue to do exactly what they do. The server adds an intelligent gateway to this, enabling your AI agent to understand what is available and when it can utilize it.
From CRM and ERP to internal knowledge bases and ticketing systems. Cube uses the MCP server to connect the sources relevant to your use case and configures which actions your AI agent is allowed to perform for each tool. No broad access to everything, but targeted integrations that unlock exactly the data needed for the tasks you want to automate.
Your data is spread across multiple systems. It is precisely in this integration that the value lies. Cube's Data Management Platform serves as a central layer to which all systems connect. The MCP server sits on top of that, allowing your AI agent to work directly with your entire data stream. Managed centrally once, rather than having to reintegrate for each application. Don’t have a DMP yet? Then we’ll build the MCP server directly onto your systems.
For organizations in the healthcare, government, and finance sectors—where data must not leave their own environment—we run open-source AI models on self-hosted infrastructure within the EU. No external connections, no U.S. cloud provider snooping on your data. We know which models are production-ready, how to manage them, and what they cost. That knowledge comes from real-world projects, not from a white paper.
For each tool on the MCP server, you can configure who has access and what actions are permitted. Every interaction can be logged. MCP supports authentication via OAuth and API keys, input validation, and rate limiting. Cube always implements solutions in accordance with ISO 27001. This means that permission management, logging, and audit trails are not an afterthought but a standard part of every delivery.
MCP is an open standard, not a product from a single vendor. If you’re using a cloud AI model like Claude, GPT, or Gemini, you can switch without having to rebuild your MCP server. If you’re running a self-hosted model on your own infrastructure, the same server will work. What you build today will still work tomorrow, regardless of which model becomes the standard.
Cube is no stranger to this field. We’ve been building mission-critical software for years. For us, AI isn’t a new playing field, but a tool we add to solutions already running in production. We run MCP servers on our own infrastructure every day. We apply that same approach to our clients’ systems: CRM packages, ERP systems, planning tools, customer portals, and internal knowledge bases. That experience is embedded in every project we deliver. We know which permission settings work reliably in production, how to design an MCP server that will still be maintainable a year from now, and where the pitfalls lie that you won’t find in any documentation.
Here’s a great real-world example: for a regional portal, Cube runs the Mistral AI model via its own MCP server. Relevant data is retrieved from connected systems and documents; the AI analyzes it and places the result in the correct location within the system. Everything remains within the EU.
Employees ask questions in plain language and receive answers based on real-time data. No need to open a dashboard or copy context. But an MCP server doesn’t just retrieve data—it draws connections. Identifying patterns in customer behavior, detecting anomalies in operational figures, making recommendations based on current information. Not just recording what happened, but looking ahead to what needs to happen. Because your AI agent has context, it provides actionable answers. Because it provides actionable answers, colleagues actually use it. Not a tool that disappears into a drawer after the demo, but an assistant that becomes part of daily work.
For each tool, you can configure who has access, which data is visible, and what your AI agent is allowed to do. Every interaction can be logged. You can restrict access to sensitive datasets. You set the boundaries, not the model. The MCP server runs on your own infrastructure. Data does not leave your environment unless you allow it to. For organizations handling personal data, subject to GDPR obligations, or dealing with sensitive business information, this is not an optional feature but a requirement. Cube runs open-source AI models self-hosted on infrastructure within the EU. No U.S. cloud provider is reading along. Everything stays where you want it to be.
Start with a single system and a single use case. Then expand step by step, without having to overhaul your architecture. Through Cube’s middleware, your MCP server works with your entire data landscape simultaneously: CRM, ERP, accounting, planning, and portals: all in a single centralized data stream. And because MCP is an open standard, you’re not locked into a single vendor. The same server works with Claude, as well as with models from OpenAI and Google.
Off-the-shelf packages are the same for everyone, so your competitive edge is exactly zero. The areas where you truly stand out are in your own processes: how you price, plan, calculate, and make decisions. The brain behind your organization. These are also the areas where AI delivers the most value, and no off-the-shelf solution fits there. An MCP server connects your AI agent to exactly that logic. Not to a generic dataset that your competitors can access just as easily, but to the rules, agreements, and working methods that make your organization unique.
Repeat
Every day, your team spends time searching for, copying, and pasting the same information from the same systems to answer the same types of questions. Each question takes five minutes to look up. Multiply that by twenty colleagues, and you’re wasting hours every day on work that a machine could do better.
Distribution
Your data is spread across six different tools, and none of them knows what the others are doing. Your CRM knows your clients, your ERP knows your orders, your project management tool knows your deadlines, but none of them has the full picture. Every decision starts with manually piecing together information from disparate sources.
Shadow AI
Your colleagues are using AI tools on their own initiative, with company data that you’d rather keep in-house. You have no visibility into where the data is going, and you don’t have a data processing agreement with half of those tools because there’s no official, secure channel.
Stagnation
Your AI pilot showed great promise, but three months later, no one is using it on a regular basis because, without context, the model can’t go beyond generic responses. The technology is there, but the connection to your data isn’t.
Growth
Your organization is scaling, and the number of systems, integrations, and data flows is growing along with it. You want to use AI as a catalyst, not as an additional source of complexity. Without a standardized MCP server, every new integration becomes a project in itself.
Compliance
You work with personal data, medical records, or financial information, and you need to be able to demonstrate who accessed which data and when. Without logging and permission management, using AI on that data is not an option. You need an integration layer that includes these features by default.
Ambition
You don’t just want AI to answer questions; you want it to get the job done: classify tickets, generate reports, and process purchase orders. You’re looking for a foundation on which to build AI agents that operate end-to-end within your systems.
C-U-B-E
We don’t start with the technology, but with the workplace: where is the manual work involved, what problem does the MCP server need to solve, and what can be automated? Together, we identify which systems are involved, where the valuable data is located, and who will be working with it. Not a months-long, in-depth discovery process, but a focused session that clarifies the scope.
Based on that, we design the MCP architecture: what tools does the server provide, what permissions are required, and how does it fit into your existing data landscape? If you already have a Data Management Platform, we’ll integrate it with that. If not, we’ll build the server directly onto your systems.
The first MCP server for a single system and a single use case is typically up and running within two to four weeks. You’ll have a working solution in no time—no need to wait months for a final product.
After going live, we monitor usage and expand the system step by step. This includes integrating additional systems, refining permissions, and adding new use cases. Your MCP server grows alongside your ambitions without requiring you to overhaul the architecture.
OpenAI is an advanced AI technology that can conduct conversations, answer complex questions and make information understandable. It helps organisations make knowledge accessible to employees and customers, turn data into actionable insights and handle routine queries automatically.
Gemini is an advanced AI platform that combines natural language and other data sources to automate and improve business processes. It can interpret information, perform analysis and perform tasks within different applications and workflows.
Claude (Anthropic) can process information and unlock knowledge without exposing or using sensitive data to train the AI model externally. This allows organisations to apply AI securely within internal tools, portals and business processes.
Microsoft Copilot is an AI assistant built into familiar Microsoft 365 applications such as Word, Excel and Teams. It helps employees automate repetitive tasks, prepare documents and reports faster and analyse data instantly.
LLaMA is an advanced language model that helps organisations understand, analyse and leverage large amounts of information.
LangChain is a framework that enables language models to be linked to data sources, applications and workflows. It enables organisations to automate complex tasks, enrich knowledge bases and internal tools with AI, and make customer portals smarter.
Supabase is an open source alternative to Firebase and runs on PostgreSQL. The platform offers everything needed for modern applications: databases, authentication, storage and APIs. In combination with AI, Supabase makes it possible to securely unlock data for smart applications, while ownership and control remain entirely with the organization.
TypeScript, a superset of JavaScript, adds static types, resulting in strong and error-resistant code. It helps detect errors early and significantly improves the development process.
Laravel, a leading PHP framework, accelerates development with tools like Eloquent ORM, Blade templating, and built-in security features. It provides an efficient and reliable foundation for building web applications.
One conversation, more clarity
A relaxed, no-obligation 30-minute chat with practical advice tailored to your situation. Mans is happy to think along with you.
With courage, pragmatism, and a focus on quality.
Amber Projectmanager
_lagekwaliteit.-2.jpg&output=webp&q=90&hash=MTc1NDI5MTI2NA==&w=540&h=660&fit=cover&a=top)
Bart Developer
Bernard Developer
Bob Test Engineer | Security & Privacy
Demi Officemanager
Dennis Full Stack Developer
Guus Full Stack Developer
Jarno Business Director
Jarno Back-end Developer
Jasper Developer
Jenne Digital Marketeer
Jeroen Back-end Developer
Job CEO | Executive
Joost Tech Lead | Team Pentagon
Jordy Software Architect
Justin Tech Lead | Team Hexagon
Kees Projectmanager
Kevin Developer
Laura Digital Designer
Maartje Content Marketeer
Mans Operations Manager
Marcus Projectmanager
Marleen Projectmanager | Team Lead Octagon
Mart Front-end Developer
Melanie Projectmanager | Team Lead Hexagon
Mick Tech Lead | Team Nexus
Nicky Projectmanager
Onno Business Controller
Remi Software Architect
Rogier Business Consultant
Rosan Manager People & Organization
Roy Full Stack Developer
Ruben Full Stack Developer
Ruben Digital Designer
Sander Developer
Stijn Front-end Developer
Tamara Business Consultant
Thomas Back-end Developer
Tom Tech Lead | Team Octagon
.jpg&output=webp&q=90&hash=MTc3MTg0MzIwMw==&w=540&h=660&fit=cover&a=top)
Wesley Projectmanager
MCP is an open standard that enables AI models to communicate in a controlled way with external systems such as CRM, ERP, or internal databases. Originally developed by Anthropic, it is now governed by the Linux Foundation.
Claude (Anthropic), models from OpenAI, Google, and Microsoft all support the protocol. A well-built MCP server works with any model that implements the standard.
Yes, when implemented correctly. MCP runs on your own infrastructure. Cube always implements with permission management, authentication, and audit logging in line with ISO 27001.
The investment depends on the number of systems, the complexity of your data landscape, and whether a central data layer is needed first. A first MCP server for one system is typically operational within two to four weeks. Contact us for a tailored estimate.
Any system with an API can be connected, and some systems already have their own MCP server. Think CRM systems like HubSpot or Salesforce, ERP packages, planning tools, customer portals, and ticketing tools. What the MCP server can do depends on what that system's API allows.
A DMP ensures data is centralised and structured. MCP is the layer that gives AI models access to that data. They complement each other.
An API is static and pre-programmed for a specific data flow. MCP is dynamic: the AI model discovers which tools are available and decides when to use them. In most organisations, both work alongside each other.
