Integrated MCP server

The integrated MCP server in NET Genium

Every NET Genium instance includes an integrated MCP server (Model Context Protocol). Through it, an AI assistant or agent – Claude or its developer variant Claude Code, ChatGPT or its developer variant Codex, Microsoft Copilot Studio or any other tool supporting the MCP protocol – connects to the instance and gets everything it needs to work with a particular implementation: the data of its applications, a description of how the applications are built, and the knowledge of the NET Genium framework.

The server is part of the installation, there is nothing to install or start separately. The procedure for connecting the individual AI tools is provided in the separate guide Connecting an AI tool to the MCP server.

What the server can do

• Working with data – reading and searching records and counting them, including the change history and attachments; with the write permission also creating, updating and deleting records.
• Describing the implementation – which applications, edit forms, view pages and rights the instance contains and how they relate to one another. The assistant is given the AI model of the implementation directly by the server, always in its current form – nothing is downloaded or pasted by hand.
• Building and changing applications – with the write permission, an agent can not only describe applications but also create them and keep changing them: application groups, applications, edit forms, view pages, controls, database queries and scripts.
• The knowledge of the framework – the assistant looks up and reads the relevant chapters of the NET Genium documentation, so it designs solutions according to the standards of the platform rather than to general habits. The knowledge base of the server is a selection from the documentation; the complete documentation is available in the AI assistant on the web.
• Performance analysis – directly from the running instance, the assistant finds out which pages, queries and scripts take the longest, how large the tables they run over are, and where indexes are missing. The usage statistics are collected only when logging to the database is turned on in the NET Genium settings.

An assistant with access to the data can therefore answer questions that the description of the applications alone cannot – for example “how many records exist”, “find a specific document” or “how did a record change over time”.

Access to the server

The MCP server is intended for the administrators of the instance. Every request signs in as a particular user and may do only what that user's permissions allow. The AI tool authenticates in one of two ways:

• With a personal MCP token – administrators generate it themselves: they click the avatar on the right side of the header and choose MCP token in the personal menu. This way works with any instance and also suits an agent running unattended.
• By signing in – in the Claude and ChatGPT applications it is enough to enter the address of the instance and sign in; no token is copied anywhere. This way is available only on instances running at the root of the web site over HTTPS, so always in NET Genium Online.

What the AI tool may do is determined by the permissions of the token, or by the consent given when signing in:

The knowledge of the framework is given to every administrator regardless of the permissions ticked – it is the general documentation of the platform, not the content of the instance.

Access to the server can be narrowed down to selected administrators only. A detailed description of this restriction is provided in the separate guide NET Genium settings.

Writing data through the MCP server

Writing through the MCP server does not run the scripts of the edit form – so neither the validations, recalculations, notifications nor status changes that follow saving a record take place. The record is created, but the process logic around it is not. Data whose saving such logic follows therefore belongs in the edit form, even on a development instance; other data can be written through the MCP server in production as well.

By default, writing is not recorded in the change history and does not take part in synchronization – which suits bulk batches, typically sample data. For changes that have to stay traceable, writing with the change history therefore has to be requested explicitly. Deleting records through the MCP server is never recorded in the change history.

In NET Genium Online the server refuses writing while the instance is in read-only mode (credit used up or data allowance exceeded).

Security

• Every request signs in as a particular user – the owner of the token, or the one who allowed the AI tool by signing in – and is subject to the allowed IP addresses and to licences just like a regular login.
• Permissions are verified on every request – losing administrator rights ends the access immediately.
• Changes to applications are subject to the same restrictions as the configuration dialogs of the admin mode – what an administrator may not change in a dialog cannot be changed through the MCP server either.
• A token is valid for 90 days and can be deleted at any time; an application connected by signing in can be disconnected at any time.
• The server is protected against token guessing.
• The write permission allows any data of the instance to be changed, not just a selected table or the user's own records – it should therefore be turned on deliberately and only for those who need it.

Typical use cases

• natural-language questions about data – “find projects without a responsible person”, “how many invoices are added each month”,
• consistency checks and summary analyses of data across applications,
• finding out when and how a record changed,
• orientation in an implementation – how the application is built and which tables and columns belong to it,
• building new applications and changing existing ones,
• performance analysis – which pages, queries and scripts take the longest and how large the tables they run over are,
• checking how a particular thing is properly done in the NET Genium framework before designing anything.

This server is naturally followed by agentic development – working with an AI agent that knows both the framework and the particular implementation and can work in it purposefully.