Ask Finland’s open data.

Aura knows 13,897 datasets from 46 sources and 394 publishers – from Statistics Finland and THL to municipal map services. Ask in English, Finnish or Swedish: you get the number, the dataset and the source it came from.

Connect your AI assistant Browse datasets

Three steps from question to rows

An AI assistant uses Aura the way a research librarian would: find the right dataset, learn what is in it, then fetch the rows straight from the publisher’s API.

  1. Find find_data

    Searches every source. Region filters understand municipalities, regions, wellbeing services counties and postcodes, in Finnish and Swedish.

  2. Understand inspect_dataset

    Returns the description, fields, licence, quality, link availability and how to query the data.

  3. Fetch query_source

    Fetches rows from the source: PxWeb, WFS, Finnish Meteorological Institute queries, OData, CSV and JSON. The exact query URL and licence come with every answer.

Connect your AI assistant

Aura speaks the Model Context Protocol (MCP). Claude, ChatGPT, Cursor and other MCP clients can use it with one setting.

Claude Code

claude mcp add --transport http aura https://aura.futuai.fi/mcp

Claude Desktop, Cursor and others

{
  "mcpServers": {
    "aura": { "url": "https://aura.futuai.fi/mcp" }
  }
}

No account needed. The service is read-only: search tools are available, tools that modify the database are not. Endpoint: https://aura.futuai.fi/mcp

For publishers and maintainers: https://aura.futuai.fi/mcp/laatu – metadata quality, missing fields and resource availability for your own datasets.

Built for AI agents

  • Every response is structured and follows a published schema (outputSchema). The agent never has to parse free text.
  • Numbers come with the source, the exact query URL, the licence and the retrieval time, so every answer can be checked.
  • Responses suggest sensible next calls (next_actions). Errors carry a code, a hint and a corrected call.
  • The public tool surface is small: five tools and instructions under 1,500 characters, so Aura does not eat the agent’s context.
  • Machine-readable service description: /llms.txt.

Who uses it

Residents
What do apartments cost per square metre in my postcode area?
Journalists
Where did unemployment rise most last year?
Researchers
Which dataset has forest basal area on a map grid?
Municipal staff
How does our population structure compare with our neighbours?
Developers
Which API gives live train positions?
Data publishers
Which of our datasets lack a description or a licence?

Browse by topic

Publishers describe their datasets with keywords (in Finnish). Each keyword leads to its datasets and to related topics.

politiikka 1591 puolueohjelma 1586 yhteentoimivuus 823 erityisohjelma 770 liikenne 633 koodisto 609 luokitus 609 vaaliohjelma 445 kunta 436 kuntatalous 426 tilinpäätös 426 kuntayhtymä 415 julkinen talous 399 Kansallinen 384 kunnallistalous 380 Alueellinen 306 paikkatietoaineisto 303 kouluterveys 281 Liikenneverkot 274 yleisohjelma 269 maa 268 Kunnat 263 Maankäyttö 255 met 207

All keywords

Aura Pro

This server runs the extended version of Aura. On top of the open version it offers:

Ready-made indicators
More than 200 indicators for municipalities, regions and wellbeing services counties in a single call: population, unemployment rate, housing prices, morbidity index. Tools get_facts, time_series and compare_areas.
Search inside the data
The index includes statistical classification values, column names of spatial datasets and code list concepts, so a term is found even when no title contains it.
Ask in any language
Vector search connects English and Swedish questions to the Finnish-language catalogue.
Honest numbers
When two statistics measure the same thing differently, the answer gives both and explains the difference.

The same approach for your own data

Aura is not tied to open data. The same structure works for an organisation’s own databases, APIs and spatial data services: a harvester per source, one metadata catalogue, search that understands Finnish, and MCP tools that cite the source of every number. An AI assistant can then find the organisation’s own data the way it now finds a Statistics Finland table.

Harvesters exist for CKAN, PxWeb, WFS, WMS, ArcGIS, OData, OpenAPI and GTFS sources, and a new source is usually one file. The quality profile shows which datasets lack a description, a licence or a working link.

The code, instructions for writing a new harvester and the discussion are on GitHub: github.com/trotor/aura.

Open source – run it yourself

Aura is MIT licensed and all of its code is public. This server differs somewhat from what you get on GitHub – Aura Pro. Your own instance runs the open version.

Search on this server also reaches inside the datasets. Indeksiin kuuluvat myös tilastotaulujen luokitusarvot, paikkatietoaineistojen sarakenimet, koodistojen käsitteet ja puolueohjelmien sanasto. Siksi haku löytää aineiston sanalla, joka ei esiinny yhdessäkään otsikossa — esimerkiksi “luomutuottajat”, “maalajieroosio” tai “kulotus”.

git clone https://github.com/trotor/aura
cd aura && python -m venv .venv
source .venv/bin/activate
pip install -e .
aura serve --http

github.com/trotor/aura. The database is part of the repository, so the datasets are there right after cloning – no harvesting needed.

This service is hosted by Futuai Oy.

To improve the service, it records search terms, requested areas, datasets and indicators, and daily tool usage counts. No user, session or IP address is stored.

Author

Aura is built by Tero Rönkkö. The project is made possible by Futuai Oy, which provides and maintains the public service so that anyone can connect Aura to their assistant without installing anything.

Contact: tero@futuai.fi. Source code: github.com/trotor/aura.

The catalogue in numbers

datasets
13,897
resources
33,435
publishers
394
sources
46

Largest publishers

Tilastokeskus
2978
Terveyden ja hyvinvoinnin laitos (THL)
2248
Yhteiskuntatieteellinen tietoarkisto
1586
Luonnonvarakeskus (LUKE)
677
Valtiokonttori
469
Paikkatiedot ja kaukokartoitus
448
Väylävirasto
380
Ulkoinen lähde: Paikkatietohakemisto
299
Kansaneläkelaitos (Kela)
275
Helsingin kaupunginkanslia
242

Most common formats

HTML
9290
CSV
4989
JSON
4764
WMS
3379
PXWEB
3047
WFS
1861
XLSX
951
PDF
417
ZIP
417
API
314
Sources (46)
Source Datasets Updated
sotkanet 3797 2026-09-27 07:23
statfin 2119 2026-10-08 14:43
avoindata.fi 1887 2026-09-27 07:12
pohtiva 1586 2026-09-27 07:22
paikkatietoikkuna 700 2026-09-27 07:21
luke 663 2026-09-27 19:38
syke 644 2026-09-27 19:39
koodistot 608 2026-09-27 07:22
hri.fi 593 2026-09-27 07:13
digitraffic 175 2026-09-27 07:19
fmi 160 2026-09-27 07:21
lipas-api 139 2026-09-27 07:21
luke-opendata 128 2026-07-26 14:13
kelasto 86 2026-09-27 10:01
traficom-tilastot 81 2026-09-27 19:39
sanastot 49 2026-09-27 07:22
valtiokonttori 48 2026-09-27 07:21
vipunen 47 2026-09-27 07:36
metsakeskus 46 2026-09-27 07:21
digitransit 40 2026-09-27 07:19
ruokavirasto 38 2026-09-28 16:10
kunnat 36 2026-09-27 07:22
traficom 32 2026-09-27 07:21
vaalirahoitus 27 2026-09-27 07:21
fluentoutdoors 24 2026-09-27 07:21
manual 16 2026-02-20 21:58
lajitietokeskus 11 2026-09-27 07:22
mml 11 2026-10-10 16:28
tulospalvelu 11 2026-09-27 07:21
luke-kartta 10 2026-09-27 07:22
statfin-geo 9 2026-09-27 19:39
finavia 8 2026-09-27 07:21
gtk 8 2026-09-27 07:21
eduskunta 7 2026-09-27 07:21
smartsmear 7 2026-09-27 07:22
overture 6 2026-09-27 07:21
ptv 6 2026-09-27 07:22
finap 5 2026-09-27 07:21
finlex 5 2026-09-27 07:21
paituli 5 2026-09-27 07:21
vayla 5 2026-09-27 07:21
taustakartat 4 2026-09-27 07:21
lipas 3 2026-09-27 07:21
prh 3 2026-09-27 19:39
kirjastot 2 2026-09-27 19:39
stuk 2 2026-09-27 07:21

Browse all datasets (in Finnish)