digna Agentic: Data Quality Answers From Your AI Assistant

Most data teams have questions they would like to ask their data in plain words: which product moved the most last month, or why a table went red overnight. Answering them usually means opening a dashboard, clicking through to the right check and explaining what it shows to whoever asked. The information is there, it just takes a person who knows where to look.

digna Agentic, new in Release 2026.11, gives that job to your AI assistant. Through the digna MCP server, Claude and other AI agents work with digna directly, so you can ask digna from your assistant and get an answer built on checks that have already run. This short video shows how it works in digna's Pharma demo project, with fictional data and the real product.

What an agent can do with digna

The digna MCP server speaks the Model Context Protocol and covers both sides of working with digna. On the configuration side, an agent can set up projects, data sources, rules, inspection schedules and subscriptions, so a new table can be brought under monitoring from a conversation rather than a series of forms. On the results side, it can read inspection statuses for a project, a data source or a dataset, all the way down to the single check, and tell you what is green, what is red and since when.

When something does go red, the agent can follow the alert the way an engineer would: from the data source to the segment that moved, and from the segment to the exact value that fell outside its expected range. Instead of a red status, you get an explanation of which part of the data changed and by how much.

Descriptions that give the data its meaning

An agent can only explain data it understands, and a column with a short technical name doesn't say much on its own. That's why digna lets you describe each data source and each column in plain language: what a row represents, what a figure measures, which unit it is in. The agent reads those descriptions alongside the results, which lets it turn a check on a column into a sentence a product manager can act on, and lets it answer business questions from what digna monitors, not only questions about data quality.

Which product changed the most?

In the demo, the question is simply: which product changed the most? The answer comes back as Amlodipine 5 mg in Czechia. Its 28-day sales fell from 1,376 packs in the 28 days to 31 May to 411 packs in the 28 days to 28 June, a drop of 70 % in four weeks. The agent also says when and where: the drop starts on 1 June, and it happens in Czechia only.

The comparison behind that answer is deliberately narrow. Seasonal products and new launches are left out, since a product swinging with the calendar or a launch ramping up would otherwise top any "changed the most" list. What the agent reports is the change digna measured, where and from when, which is the starting point for the people who know the market to find out why.

When the configuration can't answer yet

Not every question has an answer in what digna monitors today, and a useful agent says so rather than guessing. When the current configuration doesn't cover a question, the agent tells you it can't answer yet and suggests what to add, for example a new dataset. Because the same agent can also set up data sources and rules, closing that gap is one step further in the same conversation.

Running in your own infrastructure

All of this runs in your own infrastructure. digna is installed on-premises or in your private cloud, so your data never leaves your infrastructure to be monitored. You connect your AI agent to your own digna instance, and the agent works with the projects, checks and results that are already there.

The digna Agentic documentation describes what the MCP server can do, and the guide on connecting an AI agent walks through the setup.

Take the product tour to see digna Agentic and the other modules for yourself.

Frequently asked questions

What is the digna MCP server?

It lets Claude and other AI agents work with digna through the Model Context Protocol: manage projects, data sources, rules, inspection schedules and subscriptions, read results down to the single check, and investigate alerts down to the exact value.

Can an AI agent answer business questions, not just data quality questions?

Yes. With plain-language descriptions of your data sources and columns, an agent understands what the monitored data means and can answer business questions from what digna monitors, such as which product changed the most in the last four weeks.

What happens if digna can't answer a question yet?

The agent says that the current configuration can't answer it and suggests what to add, for example a new dataset. It can then help set that up through the same MCP server.

Where does digna Agentic run?

In your own infrastructure. digna runs on-premises or in your private cloud, your data never leaves your infrastructure to be monitored, and you connect your AI agent to your own digna instance. The connection guide shows how.