digna Release 2026.11: Agentic, Data Reconciliation and the Engine
|
4
min read
The questions that take a data team longest to answer are rarely the hardest ones. Does the mart still add up to the tables it was built from? Does every key still exist in the other database it refers to? Which product moved the most last month? Each usually means a query someone has to write.
Release 2026.11 adds three things to digna, the Data Quality & Observability platform, aimed at those questions: digna Agentic, digna Data Reconciliation and the digna Engine. The 75-second overview walks through all three in digna's demo projects (fictional data, real product), and the release notes list every detail.
digna Agentic: ask digna from your assistant
With digna Agentic, Claude and other AI agents work with digna through the digna MCP Server, which speaks the Model Context Protocol. An agent can manage projects, data sources, rules, inspection schedules and subscriptions, read inspection statuses down to the single check, and follow an alert from the data source to the segment and the exact value behind it.
You also describe in plain language what each data source and attribute contains, so the agent knows what it is reading, and it can then answer business questions from what digna already monitors. In the pharma demo, a fictional generics maker, the question "which product changed the most?" comes back as Amlodipine 5 mg in Czechia, whose 28-day sales fell from 1,376 packs to 411, down 70 % in four weeks, starting on 1 June and in Czechia only. If the configuration can't answer yet, the agent may suggest a change, such as adding a dataset. The digna Agentic video shows the exchange.
digna Data Reconciliation: does the target still match its sources?
This new module compares a target data source with the sources it was built from and alerts you when they no longer match. You configure the rules on the target's Data Reconciliation tab, and any data source in the project that has reconciliation enabled can serve as a source.
An exact rule works like a data quality rule for one source and one target: you pair checks by statistic and attribute, such as row counts or amount sums, and set an absolute or relative tolerance, and its timeline shows each day's difference against it. In the retail demo, a daily sales aggregate showed 51,747 euros in gross revenue on 26 August, while its own sales lines summed to 66,205 euros: 21.8 % short, against a tolerance of ± 0.01 %.
An AI rule combines checks from several sources with one target, with no joins, filters or aggregations to define. A model learns from history how the target follows its sources and alerts when the target falls outside the band of the rule's past errors; each rule has Sensitivity, Memory and Model Complexity settings. In the demo, an order mart built from ten source tables held 1,742 orders on 6 October, where its sources predicted about 2,771. Every reconciliation check appears on the Dashboard, counts towards the inspection status and triggers notifications. The digna Data Reconciliation video covers both rule types.
The digna Engine: choose where profiling runs
Every database connection now has a profiling location, Source Database by default or the digna Engine, for every supported source technology. The release also brings cross-connection referential integrity checks to digna Data Validation, so one check can span two connections and technologies, for example Oracle and PostgreSQL, Teradata and Snowflake, or Databricks and SQL Server. When profiling stays in the source database, only the key columns are exported to the Engine, and the Dashboard shows when a check runs across two connections. The digna Engine video shows the setting.
Smaller additions to profiling
Query attributes let you define a column as a SQL expression and have it profiled like any other, there is a new Date column category, and you can profile a single attribute on demand against one date's snapshot.
Who benefits
Data engineers can validate ETL results, finance and controlling teams can run standard reconciliations as exact rules, governance teams can check referential integrity across technologies, and business users can ask digna through an agent. All of it runs on-premises or in your private cloud, so your data never leaves your infrastructure. To see them in context, take the digna product tour.
For a closer look at the new module, including both rule types and the retail and order-mart examples, see how digna Data Reconciliation checks that targets still match their sources.
Frequently asked questions
What is new in digna Release 2026.11?
digna Agentic, which lets AI agents work with digna through the digna MCP Server, the new digna Data Reconciliation module, and the digna Engine, with a per-connection profiling location and referential integrity checks across connections. Profiling also gains query attributes, a Date column category and on-demand profiling.
Which AI agents can work with digna?
Claude and other AI agents, through the digna MCP Server and the Model Context Protocol. They can manage configuration, read results down to the single check, investigate alerts and answer business questions.
What is the difference between exact rules and AI rules in digna Data Reconciliation?
An exact rule pairs a check on one source with a check on one target and compares them against an absolute or relative tolerance. An AI rule combines checks from several sources, learns how the target follows them and alerts when it falls outside the band of the rule's past errors.
Can digna check referential integrity between two different databases?
Yes. Referential integrity checks in digna Data Validation can span two connections and technologies, such as Teradata and Snowflake.
