Every source, one model, dashboard-ready.
Your numbers live in five places, and classic tools want them clean and in one place before you can start. Agentigator does not. Connect straight to analytics-ready data: a data mart, a warehouse, CSV or Excel, and a folder of same-format Excel files reads as one source, natively. No pipeline needed. When data really does live in several systems, or needs cleaning first, the ETL is already built in: connect the systems you run, shape and combine them with composable steps, and build multi-fact-table models in clicks. No pipeline team, no SQL required, and full SQL the moment you want it. All of it on your own infrastructure.
The systems you already run.
Agentigator connects directly to the databases, data warehouses, data marts and files SMEs actually run, and every connection stays inside your network. Analytics-ready data needs no pipeline at all: point a dashboard at it. Credentials never leave your servers.
Read any table or view directly, with schema browsing and a live preview of the data.
Paste a SQL query and use its result as the flow input, when the extraction logic already exists.
CSV and Excel files are read directly. Several Excel files that share the same format are read as a single source, natively, no merging step required.
Composable steps, validated before they run.
Flows are built from typed steps that chain into a pipeline. Every step validates its configuration before anything runs, and previews its output, so you see the data at each stage.
Read a table or a view, or paste any SQL query. Both modes preview live before the flow runs.
Keep, drop, rename and reorder columns.
Visual conditions with nested AND and OR groups, typed values, between and list operators.
Add derived columns from a function library, or write a custom expression when you want to.
Combine two flows with left and right prefixes and explicit column selection, so name collisions never surprise you.
Enrich rows from a reference table, with a clear policy when no match is found: null, a default value, or stop.
Stack compatible flows into one.
Group and summarize: sums, counts, averages, min and max per group.
Order by any columns, ascending or descending.
Keep one row per key, deterministically.
Convert types with an explicit on-error policy instead of silent corruption.
Split a flow into branches by condition, each branch continuing its own path.
Bucket long tails into readable categories, everything else lands in a group you name, such as Others.
Produce a complete calendar table for any range.
Densify a sparse time series so charts never skip a day.
Write the way each table needs.
Two target families: Delta Lake tables in the local warehouse, or a relational database in its own dialect. Both support the same write modes, so the choice is about destination, not capability. Targets control schema evolution explicitly, so a new upstream column is a decision, not an accident.
Insert new rows, keep everything already there. The natural mode for event and snapshot feeds.
Give the keys, Agentigator updates matching rows and inserts the rest.
Full replace of the target, for reference tables rebuilt each run.
Replace only one slice, a partition on Delta or a predicate on a database, and leave the rest untouched.
Type-2 versioning built in: every change becomes a dated version, current rows flagged, past preserved.
Slowly changing dimensions, as a checkbox.
Choose the SCD2 write mode, and Agentigator does the rest: it closes the previous version, opens the new one with its validity dates, and keeps the current row flagged. The contract is identical on Delta and on database targets, and it is enforced before the flow runs, not discovered after.
Choose the columns that identify a record.
Choose the columns whose changes should create a new dated version.
The kind of table that usually takes a senior SQL developer an afternoon of window functions exists here in two clicks.
Flows, pipelines, and a scheduler that runs them.
A data flow is the unit of loading: one source, shaped and written to its target. Pipelines orchestrate many flows into a single ordered run, and the scheduler keeps everything current on its own.
The unit of loading: read a source, transform it, and write the result to its target. Self-contained, and previewed before it runs.
Compose flows into a pipeline that runs as one unit, in order, with a clear status for each flow.
Schedule automatic refreshes, so your tables stay current without anyone pressing Run.
Load only what changed instead of the whole table, for fast refreshes on large feeds.
Every load is recorded with its status, so you can see what ran, when, and whether it succeeded.
- Sales ordersSuccess
- Product lookupSuccess
- Warehouse writeRunning
- Refresh martsWaiting
Intelligence and safety, by default.
AI join suggestions
Agentigator inspects both sides and proposes the natural join keys.
Quarantine
Rows that fail a step are captured with their reason. The flow keeps its contract, nothing disappears silently.
Business names
Delta tables carry readable names, so the warehouse stays understandable by humans.
Fast local engine
Transformations execute on an embedded analytical engine, on your hardware.
All of it, on your own servers.
Every source, every transformation and every write runs on your infrastructure. Your data never leaves your walls, and the analytical engine runs on your hardware.