A platform an AI agent can actually operate
NET Genium is a low-code platform for enterprise applications. What sets it apart is not that it has an AI feature bolted on — it is that an AI agent gets real access to the whole thing: it builds applications from a description, works with the data and the application model of a running instance, and knows the framework it is building on.
Every instance runs on its own database, so there are no limits on the number of records — the only ceiling is the storage space of your plan.
Describe the application and it gets built
A new instance is not an empty screen you have to figure out. Right on the NET Genium home page you describe in your own words what you need to keep track of and how the process works in your organization.
From that description the platform builds a working application — edit forms with the matching controls, the database tables behind them, and view pages for working with the records. It is not a preview or a proposal on paper: you open the application straight away and start putting data into it.
You can have as many applications built this way as you need, and you keep developing each of them with the ordinary tools of the platform — adding fields, changing the layout of forms, adding process logic and permissions. You are never locked into what the AI generated.
Connect your own agent
Every instance contains an integrated MCP server (Model Context Protocol), so you can connect your own assistant to it — Claude, Codex, Microsoft Copilot Studio or any other tool that supports the protocol. There is nothing to install and nothing to start separately.
Through it the agent is given three things a generic connector cannot give it:
- the data of your applications — it reads the records, counts them and looks into the change history; with the write permission it changes them as well,
- the model of your implementation — applications, forms, controls, permissions and the data model, so it knows the actual shape of your solution instead of guessing,
- the knowledge of the framework — the documentation and conventions of the platform, so that what it proposes follows the current standards rather than general habits.
That is the difference between an assistant that talks about your data and one that works in your system.
The setup takes a few minutes and is described in the guide Connecting an AI tool to the MCP server; what the server offers and where its limits are is described in The integrated MCP server.
Under your control, not beside it
Access is governed by a personal token of a particular user. Every request therefore logs in as that user and is subject to the same permissions, allowed IP addresses and auditing as a regular login — the agent never sees more than its owner does.
Reading and writing are separate permissions. An agent that is only meant to answer questions and analyse data is given reading alone and cannot change anything.
Building applications with an agent
Beyond working with data, the agent also works with the application model itself. Through the same MCP server it reads the model of your implementation and works its changes into it — it sets up edit forms and view pages, adds controls, database queries and scripts, and can propose a data structure from a requirements analysis.
That is what separates it from an assistant that merely generates code you then have to transfer by hand. The agent works against the running instance, and because it takes the conventions of the platform and the shape of your solution from the same server, it follows the established structure of the project instead of introducing its own.
Every change stays under the developer's control — it arises as a specific, reviewable modification and is uploaded into the instance only once it is approved.
A detailed description is provided in the separate guide Agentic development.
A real database, not a limited base
The reason an agent can do all of this is that there is a real database underneath, not a spreadsheet with rows.
- Each instance has its own database with a real SQL data model — tables, columns, data types, relations and indexes.
- No limits on the number of records, neither per table nor per instance; the only ceiling is storage space.
- The number of applications, forms and users does not change the structure — the same platform runs solutions with dozens of applications and hundreds of forms.
An agent that is given a real data model asks better questions and makes fewer mistakes than one that is given a narrow API over a capped table.
What it costs
With a single user the instance is free of charge, with no time limit. Billing starts with the second user and no payment card is needed to start.
The details of the pricing, the data limits and the credit are described on the page NET Genium Online.
Why you can rely on it
NET Genium is not a new project. It is a platform that companies have been running for years in daily operation — hundreds of implementations and tens of thousands of users, with dozens of applications and millions of records in a single instance.
In the Czech Republic the platform runs several of the state's significant information systems, which are regulated under the national Cybersecurity Act. The instances of NET Genium Online run in data centres within the European Union.
Try it
One user free of charge with no time limit, no payment card and no installation. Your instance is created within a few minutes at identifier.netgenium.online.