Coverage and attribution

Listings and single notices carry a coverage object that describes the answer it came with. MCP tool results carry an object of the same shape.

{
  "sources": ["TED"],
  "attribution": [
    {
      "source": "TED",
      "text": "Uppgifter: Tenders Electronic Daily (TED), Europeiska unionens publikationsbyrå | Bearbetning: Fentlig AB"
    }
  ],
  "dataAsOf": "2026-08-27",
  "isComplete": false
}

sources names the sources of the rows in this response. dataAsOf is when the data was last updated. note, when present, states isComplete in words.

Where the notices come from#

Notices come from the EU's Tenders Electronic Daily (TED) and from the Swedish platforms where buyers advertise. Notices from those platforms are listed with the source Fentlig.

Notices are in Swedish and are not translated. Search in Swedish.

The default market is Sweden, meaning notices from buyers in Sweden. On the REST API, country and foreign_buyers widen it; see Getting started.

The published set is a subset#

isComplete is false. Not every procurement advertised in Sweden is in the published set. An empty result means nothing published here matches, not that nothing exists. When you show a count or a total, say that it covers the published set.

What is published#

The notice as advertised: title, buyer, CPV classification, region, procedure, deadline, estimated value, and a link to the original. Once a procurement is concluded, the outcome at market level: award date, number of bids, awarded value.

The names of winning suppliers, the text of the tender documents and the requirements extracted from them are not in the published set. Fentlig customers read them in the app and through the MCP server, according to their plan. See Authentication for which plan includes each.

Contact details are removed from text written by a buyer. Names can remain, so treat description as text that may name people.

Attribution#

coverage.attribution is a list of lines, one per source. If you republish, display or redistribute the data, reproduce every line verbatim, dates included.

Read the lines from each response rather than copying them into your code: the dates in them change. Render all of them.