AI that uses your systems safely.
AI assistants such as Claude and ChatGPT only become truly useful when they can reach your own data: the CRM, the planning, the documentation. With the Model Context Protocol (MCP) you make that connection, but only the way you want it.
MCP is an open standard that lets an AI assistant use tools and data from other systems. An MCP server determines which actions are available, for example looking up a customer, retrieving a quote or creating a ticket. The assistant cannot do anything outside that.
We design which actions are useful and responsible, build the MCP servers and take care of security: who may do what, which actions require approval and what is logged. That way your team benefits from AI on your own data, without giving up control.
Components that make the difference.
Every project is custom work. These are components we often build within this service.
Custom MCP servers
Connections to your CRM, ERP, planning, knowledge base or database, precisely scoped.
Access control
Decide per user and per role which actions and data are available.
Approval steps
Sensitive actions, such as sending or deleting something, only after confirmation by a person.
Audit logging
Every call recorded: who, what, when. For control and accountability.
Protection against misuse
Measures against prompt injection and data leaking through the assistant.
Advice and training
Which processes are suitable, and how your team works with them safely.
For organisations like yours.
- Organisations with a lot of internal knowledgeTeams that want fast answers from documentation, case files and systems.
- Sales and service teamsEmployees who want customer information and status directly in their assistant.
- IT departmentsTeams that want to allow AI, but with policy, logging and control.
What that looked like.
A few systems we built within this service. Read more about each project in the tour.
About MCP.
Is your question not listed? Email bastiaan@carescript.nl.
What exactly is MCP?+
The Model Context Protocol is an open standard that lets AI assistants use tools and data from other systems in a structured way. Read more in our knowledge base article about MCP.
Which AI assistants work with MCP?+
More and more assistants and developer tools support MCP, including Claude and ChatGPT. We advise which one fits your organisation and policy best.
Is it safe to give AI access to our systems?+
That depends on how it is set up. We only give the assistant access to clearly scoped actions, with permissions per role, approval for sensitive actions and full logging.
How quickly can we have a first connection?+
A first scoped connection, for example read-only access to your CRM, is often in place within a few weeks. After that we expand step by step.
Curious what this means for you?
Tell us briefly what you have in mind. We will plan a no-obligation call and are happy to show you a demo.

