Skip to content
iAgro

iAgro Extract

Your data already exists. It just isn't arriving.

The ingestion engine built for agriculture. It fetches data wherever it lives —spreadsheets, ERPs, machinery, agencies— and deposits it raw, without interpreting it. It is the first piece of Data Platform and works on its own.

28

sources with a native connector

10

table formats, from CSV to ISOXML

7

machinery brands

12

agencies and public data sources

The essentials

  • Discovers on its own what each source contains
  • Incremental loads that identify the file by content
  • Alerts when a source changes structure, without stopping the load
  • Detects silent failures: truncated pages, changed units, partial responses
  • Read-only credentials, never writes to the source
  • Writes to PostgreSQL, Snowflake, or S3

The typical failure is not an error: it is silence

When an integration breaks with an error, someone finds out. The real problem is the other one: the source responds “all good” and brings back half. A page that got cut off, a value that came in a different unit, an agency that returns a success code with an empty result.

A generic connector doesn't detect that, because detecting it requires knowing the source. And by the time it's discovered, there are already weeks of decisions made on incomplete data.

How it works

Discovers on its own. It goes through every source and describes what it contains, even if no one has documented it. AI helps classify, but does not decide: only what a person marked for production is sent, and human classification always wins.

Brings only what is new. It identifies each file by its content, not by its path: moving or renaming does not trigger a reload or a new analysis.

Survives changes. If a source adds a column, the load continues and notifies you. It does not break, and it does not continue as if nothing happened either.

Monitors itself. It monitors the connections and sends alerts by email, Slack or another channel. The only automatic action is retrying: pausing a connection or changing a credential is decided by a person.

The sources it connects

A generic connector treats every source the same way, and that is where data gets lost. Extract has one connector per source, and it knows its quirks: agencies that answer “all good” and return less than they should, pagination that does not follow the standard, CUIT formats —the Argentine tax ID— that must match digit by digit. Agencies in other countries come in through their open data portals or with a custom connector.

Database connectors are tested end to end, and those for SENASA, IBGE, U.S. Drought Monitor, CKAN and Socrata portals, NASA POWER, SoilGrids and Eiwa, against the live service. Google Drive and Google Sheets, OneDrive and SharePoint, John Deere Operations Center, Climate FieldView, ARCA, SIMA, ICAR ADE, SIO Granos, USDA NASS and Finnegans gO! are built on each provider's official API or specification and are fully verified with your credentials or agreement during implementation.

  • Spreadsheets and filesGoogle Drive, OneDrive, SharePoint, Amazon S3, Google Sheets, multi-sheet Excel, CSV and JSON
  • DatabasesPostgreSQL, MySQL, SQL Server, Oracle and MongoDB, always with a read-only user
  • Management systemsFinnegans gO! with a dedicated connector, REST APIs with five authentication methods and five pagination methods, and custom ERPs, scales and livestock software
  • Argentine agenciesARCA —the tax agency, formerly AFIP— with five services: grains, livestock, dairy, tobacco and electronic consignment note; SENASA —agricultural health— with the RENSPA registry of establishments and their polygons by CUIT; SIO Granos —the official registry of grain operations— with operations and prices per ton; the registry of meat and dairy operators; and Datos Abiertos del Agro
  • Machinery and agronomyJohn Deere Operations Center, Climate FieldView, Eiwa with agronomic layers per field, SIMA for field monitoring, ISOXML —the standard format of machinery monitors— from seven brands, and ICAR ADE —the dairy data standard— with 42 resources
  • Environment and soilNASA POWER for daily weather and SoilGrids for soil properties by coordinate
  • Public and international dataCKAN and Socrata open data portals from any country, IBGE from Brazil with municipal production since 1974, USDA NASS by county since 1866 and the U.S. Drought Monitor with weekly maps since 2000

Where it deposits the data

Extract writes to PostgreSQL, Snowflake, or Amazon S3, in a landing zone owned by your company. DataRefinery then reads from it to bring everything into the canonical model; in Postgres, your team also queries it with its own tools.

Why you can trust it with access

Extract enters the systems where your information lives, so limits matter as much as capabilities.

Credentials are read-only and it never writes to the source. It does not leave metadata files lying around in your folders: what it needs to know about each file lives inside Extract. The audit log is append-only, also for internal users.

And there are two things AI cannot do: correct a value and change a unit. If a piece of data arrived wrong, it stays wrong and is flagged — it is not fixed silently.

Combines with

Request a demo or start with an assessment

In a demo, we show you the platform running with test data. If you prefer to start with your own case, we run a no-obligation assessment and tell you where it makes sense to begin.