Global Tender Monitor — UK & EU Public Procurement

Watch UK and EU public tenders in one feed and get only the notices published since your last run. Buyer, contract value, deadline, CPV codes and a direct link, normalised into one schema across both sources.

Official government APIs. No API key, no scraping, no proxy needed.

Sources

Source	Endpoint	What it gives
UK Find a Tender	OCDS release packages	Contract value and currency, buyer, status, deadline
TED (European Union)	Expert search API	322,000+ notices across all member states

Output

Both sources produce the same fields, so a CSV export stays rectangular and you can sort UK and EU notices together.

{
  "source": "uk",
  "sourceName": "UK Find a Tender",
  "noticeId": "ocds-h6vhtk-05a1b2",
  "title": "Assertive Outreach Rough Sleepers",
  "buyer": "Caerphilly County Borough Council",
  "country": "GB",
  "valueAmount": 2400000,
  "valueAmountGross": 2880000,
  "currency": "GBP",
  "status": "active",
  "publishedDate": "2026-09-14T13:45:47.000Z",
  "deadline": "2026-10-16T11:00:00.000Z",
  "cpvCodes": ["85300000"],
  "url": "https://www.find-tender.service.gov.uk/Notice/..."
}

Three things handled properly

TED is multilingual and not always in English. Every TED text field arrives as a language map — {"eng": "...", "fra": "...", "hun": "..."} — and some notices carry no English at all. A preferred language is chosen, English is the fallback, and whatever the notice actually has is the last resort. Returning null because English was missing would silently drop real tenders.

IDs are namespaced by source. UK and EU both use plain numeric identifiers, so tracking them in a shared list would let one source's notice hide behind the other's. State is keyed source:id, with a test that asserts two notices numbered 100 both survive.

The first run shows you the current notices. It delivers up to maxItemsPerSource per source and remembers them. Every later run returns only what is new since the run before. RUN_SUMMARY says exactly what was delivered.

Input

Field	What it does
sources	uk, eu, or both.
mode	new returns only unseen notices. all returns everything each run.
keywords	Keep only notices whose title, description or buyer matches.
tedQuery	TED's own expert-search syntax. It rejects a bare wildcard.
language	Three-letter code for picking from TED's language maps.
stateKey	One key per watchlist so separate schedules do not interfere.

TED query examples

notice-title ~ (software)
classification-cpv IN (72000000)
buyer-country = NLD

Example: daily watch for IT contracts

{
  "sources": ["uk", "eu"],
  "mode": "new",
  "keywords": ["software", "IT services", "cloud"],
  "tedQuery": "classification-cpv IN (72000000)",
  "stateKey": "it-contracts"
}

Schedule that daily. The first run returns the current notices; every run after returns only newly published ones.

Notes and limits

• Contract values are published by the UK source. TED does not include a value in this response shape, so valueAmount is null for EU rows rather than estimated.
• A rejected TED query returns its error message directly instead of being retried, since a syntax error will not fix itself.
• If one source fails, the other still delivers and RUN_SUMMARY records the reason.
• Deleting the named state store resets the baseline, and the next run records a fresh one.

Development

npm install
npm test
node src/main.js