Uruguay Open Data Scraper - CKAN Datos Abiertos

Uruguay Open Data Scraper - CKAN Datos Abiertos API

Scrape every one of the 2,702 datasets on catalogodatos.gub.uy, Uruguay's national open data portal, in three requests. Every row carries the full catalogue record: publisher, licence, themes, tags, declared update cadence and each downloadable file with its format, size and direct URL. No login, no API key, no registration. Export to CSV, JSON, Excel, or XML.

The portal's own CKAN API is open but awkward: it caps a page at 1,000, pages its organization list at 25 whatever you ask for, spells one file format 48 different ways, and answers an unknown filter field with zero rows instead of an error. This reads all of it, normalises it, and adds what the API will not tell you on its own: the real column schema behind every queryable table, and whether the download links still work.

Who uses it	What they scrape Uruguay open data for
Data journalists	Which ministries publish what, and which datasets have gone stale
Civic tech and NGOs	A complete, refreshable index of Uruguayan public data
Researchers and academics	Statistics, health, housing and labour datasets with their real column schemas
Data engineers	A machine-readable manifest of 14,373 files to feed an ingestion pipeline
Open data auditors	Link rot, missing licences, overdue update cadences, harvest provenance

What it does

This Actor reads Uruguay's CKAN catalogue and returns it as flat rows, in ten different shapes chosen by one Output setting. Point it at the whole catalogue, at a filtered slice of it, or at a list of datasets you name by URL, slug or UUID. Every dataset row carries:

• 🧾 Core fields: title, description, slug, portal URL, publishing organization with its own description and logo, author and maintainer with emails, licence, version, and creation and modification dates.
• 🗂️ Classification: the 22 portal themes, the full tag list, and every file format present, deduplicated.
• 📦 Files nested inline: each one with format, MIME type, byte size, checksum, direct download URL, portal URL and dates.
• 🧮 Computed signals: days since last update, the declared cadence in plain language, an overdue flag, total bytes, and whether the dataset has queryable DataStore tables.
• 🔎 Provenance: whether the record was harvested from another portal, and which one.

Results export to CSV, JSON, Excel, or XML, or stream from the API.

What you can do with Uruguay open data

🗺️ Build a complete index of the catalogue.

One run with no filters returns all 2,702 datasets in three requests. Schedule it and diff the updatedAt column to see exactly what the state published or changed this week.

🔬 Find the data before you download it.

Turn on the column schema block and the Actor reads each DataStore table's real column names, PostgreSQL types and exact row count without downloading a byte. Across just 25 popular datasets that surfaced 101 live tables holding 6.5 million rows, including a vote breakdown with 519,750 rows and 8 typed columns.

🔗 Audit link rot and staleness.

The portal harvests six external sites and never re-verifies what it copied. Switch on the link check and every file URL is fetched and reported with its HTTP status. Sampling 90 URLs across the full age range of the catalogue, 8 were not retrievable.

📇 Map who publishes what.

The Contacts output deduplicates the people and units named on the datasets into a directory with emails, roles, dataset counts and the themes they cover. 316 distinct contacts across the whole catalogue.

Why choose this scraper

	What you get
Ten outputs, one Actor	Datasets, files, DataStore tables, contacts, organizations, themes, tags, formats, licences and harvest sources.
Real column schemas	Column names, PostgreSQL types and exact row counts read from the portal DataStore, with no download.
Link checking	Every file URL fetched and reported with status, content type, size, redirect and latency.
Filters measured, not guessed	Every filter in the schema was tested against the unfiltered total of 2,702, and only verified values are offered.
Format names that work	One canonical CSV matches all nine spellings the publishers actually typed, and finds 2,550 datasets instead of 2,543.
Four export formats	CSV, JSON, Excel and XML, from the dashboard or the API.

How it compares

The generic CKAN scrapers on the Store point at any portal and return the raw package record, which is a reasonable thing to want and cheaper per row. This one is built for this catalogue specifically, so it knows the things a generic reader cannot: that update_frequency is a Uruguay-only field holding a number of days, that the organization list silently pages at 25, that include_tracking breaks the search endpoint, and which four of the six harvest sources are dead test entries. It also does work the others do not attempt at all, reading DataStore schemas and checking whether the files are still there. If you only need raw CKAN packages from any portal, the generic ones are the better buy.

Feature	ParseForge	straightforward_hydra	datapilot	doggo
Raw CKAN dataset record	Yes	Yes	Yes	Yes
DataStore column schema and row counts	Yes	No	No	No
Sample data rows from the live table	Yes	No	No	No
File link checking	Yes	No	No	No
Verified Uruguay filters and enums	Yes, 68 publishers, 22 themes	Generic	Generic	Generic
Separate outputs beyond datasets	10	1	1	1
Billable events	19	2	2	2

What a dataset looks like

Every dataset returns as one JSON row with its files nested. Here is a real record from a verified run. Three things are shortened for space and nothing else is changed: the description is truncated, the publisher's own long description is dropped, and only the first of its four files is shown, with that file's portal URL and position fields left out.

{
  "datasetId": "accd0e24-76be-4101-904b-81bb7d41ee88",
  "name": "inumet-observaciones-meteorologicas-temperatura-del-aire-en-el-uruguay",
  "title": "Observaciones meteorológicas - Temperatura del aire en el Uruguay",
  "url": "https://catalogodatos.gub.uy/dataset/inumet-observaciones-meteorologicas-temperatura-del-aire-en-el-uruguay",
  "apiUrl": "https://catalogodatos.gub.uy/api/3/action/package_show?id=inumet-observaciones-meteorologicas-temperatura-del-aire-en-el-uruguay",
  "description": "El siguiente conjunto presenta la información sobre las observaciones meteorológicas llevadas a cabo por estaciones meteorológicas automáticas del Ins …",
  "organizationName": "inumet",
  "organizationTitle": "INUMET",
  "organizationUrl": "https://catalogodatos.gub.uy/organization/inumet",
  "author": "INUMET",
  "authorEmail": "contacto@inumet.gub.uy",
  "maintainer": "Área de Informática",
  "maintainerEmail": "informatica.direccion@inumet.gub.uy",
  "licenseId": "odc-uy",
  "licenseTitle": "Licencia de DAG de Uruguay",
  "isOpenLicence": "Yes",
  "sourceUrl": "Not Disclosed",
  "version": "1.0",
  "categories": ["medio-ambiente"],
  "categoryTitles": ["Medio Ambiente"],
  "tags": ["Aire", "Cambio climático", "Medio ambiente", "Meteorología", "Temperatura"],
  "tagCount": 5,
  "resourceCount": 4,
  "formats": ["CSV", "JSON", "XLSX", "XML"],
  "totalSizeBytes": 63031037,
  "hasDatastore": "Yes",
  "datastoreTableCount": 2,
  "createdAt": "2024-10-22T17:59:12.802088",
  "updatedAt": "2026-08-27T04:15:14.268125",
  "daysSinceUpdate": 0,
  "updateFrequencyDays": 1,
  "updateFrequencyLabel": "Daily (Diaria)",
  "updateOverdue": "No",
  "harvested": "No",
  "harvestSourceTitle": "Not Disclosed",
  "resources": [
    {
      "resourceId": "2435e1f5-24c7-4991-8eb7-7708b556b32f",
      "name": "Metadatos de temperatura del aire",
      "format": "JSON",
      "mimeType": "application/json",
      "downloadUrl": "https://catalogodatos.gub.uy/dataset/accd0e24-76be-4101-904b-81bb7d41ee88/resource/2435e1f5-24c7-4991-8eb7-7708b556b32f/download/metadatos_inumet_temperatura_del_aire.json",
      "sizeBytes": 1319,
      "checksum": "Not Disclosed",
      "queryableApi": "No",
      "datastoreApiUrl": "N/A",
      "createdAt": "2025-08-30T04:17:03.548584",
      "lastModifiedAt": "2026-08-27T04:07:51.942993"
    }
  ],
  "scrapedAt": "2026-08-27T16:21:23.929Z"
}

Fields the portal leaves blank come back as Not Disclosed, and questions that do not apply to a row come back as N/A, so an empty CSV cell always means the same thing. sourceUrl is empty here because this publisher did not fill it in.

Configure the run

Pick an Output first: it decides what a row is and which event you pay. Then either name the datasets you want in Dataset URLs or slugs, or leave that empty and narrow the whole catalogue with filters, which combine with AND while multiple values inside one filter combine with OR. Leave everything empty and you get all 2,702 datasets. Optional blocks are all off by default and each is billed only when it actually returned something. The Input tab lists every parameter.

Pull three named datasets, given as a portal URL, a bare slug and a UUID:

{ "dataset": "datasets", "startUrls": ["https://catalogodatos.gub.uy/dataset/agesic-guia-de-tramites", "montevideo-audiovisual", "accd0e24-76be-4101-904b-81bb7d41ee88"], "maxItems": 10 }

Health and environment datasets from three ministries, only ones with live queryable tables, with their column schemas read:

{ "dataset": "datasets", "categories": ["salud", "medio-ambiente"], "organizations": ["msp", "ambiente", "inumet"], "onlyWithDatastore": true, "includeDatastoreSchema": true, "maxItems": 200 }

Every CSV file in the catalogue, checked to see whether it still downloads:

{ "dataset": "resources", "formats": ["CSV"], "checkResourceLinks": true, "onlyLiveResources": true, "maxItems": 5000 }

Free users

Free-plan runs return up to 10 rows as a preview. Upgrade your Apify plan to collect up to 1,000,000 rows per run.

Run it

1. Create a free Apify account with $5 in credit.
2. Open the Uruguay Open Data Scraper.
3. Pick an Output, add any filters and optional blocks, set Max Items, and click Start.
4. Export the results as CSV, Excel, JSON, or XML from the Dataset tab.

Run it programmatically through the Apify API or the ApifyClient for JavaScript and Python.

Use with AI agents (MCP)

Give an AI agent live access to Uruguay's open data catalogue through the Model Context Protocol. Add the Actor to Claude, Cursor, or any MCP client:

claude mcp add --transport http apify "https://mcp.apify.com?tools=parseforge/uruguay-open-data-scraper"

Then prompt it in plain language:

• "List every health dataset the Uruguayan government publishes and say which ones are overdue for an update."
• "Find Uruguayan open datasets that have a queryable table with a date column, and show me the column names."
• "Which Uruguayan ministries publish the most datasets, and what formats do they use?"

Copy this into ChatGPT, Claude, or Cursor to start:

Use the Apify Actor "parseforge/uruguay-open-data-scraper" to read Uruguay's national open data catalogue at catalogodatos.gub.uy. Input: { "dataset": "<datasets|resources|datastore-tables|contacts|organizations|categories|tags|formats|licenses|harvest-sources>", "searchQuery": "<text>", "organizations": ["<slug>"], "categories": ["<theme>"], "maxItems": <n> }. It returns title, description, organization, licence, themes, tags, update cadence and every file with its format, size and download URL. Call it with the ApifyClient and my APIFY_TOKEN.

Troubleshooting

Why am I getting no results?

Check your filters first. If you used Filter query, note that this portal answers an unknown Solr field name with zero rows rather than an error, so a typo looks like an empty catalogue. Tag values must match the portal's own spelling exactly, accents and capitals included: run the Tags output to see them.

Why fewer rows than I asked for?

Only datasets with queryable tables and Exclude formats are applied after the search, so they remove rows the index already returned. The run over-fetches to compensate, but it cannot invent datasets that do not match. Free-plan runs also stop at 10.

Why is a field empty?

It is not empty, it says Not Disclosed or N/A. The first means the publisher left that field blank, which is common for sourceUrl, version and checksum. The second means the question does not apply, for example a DataStore API URL on a file that has no table.

Why is the run slow?

The catalogue sweep itself is fast: 14,373 file rows in three requests and 43 seconds. The optional blocks are what cost time, because each one is a separate request per file. Turning on link checking, schemas, previews and visualizations together means four extra requests per file. Turn on only what you need, or lower Max Items.

A field stopped filling.

The portal changed its CKAN response shape or an endpoint moved behind its edge gateway. Email us with your run ID so we can update the parser.

FAQ

Question	Answer
Do I need an API key or an account on the portal?	No. The CKAN API is fully anonymous and needs no key, cookie or proxy.
How is this different from calling the CKAN API myself?	It handles the traps: the 1,000-row page cap, the organization list that pages at 25, 48 raw spellings for 30 formats, and filters that fail silently. It also adds DataStore schemas and link checks, which CKAN does not expose in the search response.
How many datasets are there?	2,702 datasets and 14,373 files, from 68 publishing organizations, measured 2026-08-27.
What is a DataStore table?	A file the portal has loaded into its own database, so it can be queried and filtered live. 101 of 400 files in a sample run had one. The Actor returns their column names, types and exact row counts.
Can I get the actual data, not just the metadata?	Yes, for DataStore-backed files: switch on sample rows and you get the records themselves, verbatim, up to 100 per table. For everything else the row carries a direct download URL.
Are the download links reliable?	Not always. Of 90 file URLs sampled across the catalogue's full age range, 8 did not answer. Switch on link checking to find out which.
Is the catalogue in Spanish?	Yes. Titles, descriptions and tags are Spanish. Theme names are returned in both Spanish and English.
Can I filter by date?	Yes, on both the metadata creation and modification dates, as YYYY-MM-DD.
Can I scrape specific datasets rather than search?	Yes. Put their portal URLs, slugs or UUIDs in Dataset URLs or slugs, mixed freely. It works with every Output, so the same list can give you their files or their DataStore tables.
How fast is it?	335 rows a second on the file output, and the whole 2,702-dataset catalogue in three requests. Optional blocks add a request per file and slow it down proportionally.
Is this an official Uruguayan government product?	No. It is unofficial and reads only public data from the portal's public API.

🆘 Need help? Email parseforge@protonmail.com with your run ID, your input, and what you expected.

⚠️ Disclaimer. This Actor is unofficial and is not affiliated with, endorsed by, or sponsored by AGESIC or the government of Uruguay. It collects only publicly available data from the portal's public API. You are responsible for using the data in compliance with the portal's terms and applicable laws, including GDPR, CCPA, and PIPL. Contact fields come from the datasets' own public metadata; do not use them to identify, profile, or target individuals.