Public Tenders Feed: EU, UK, US, Canada, New Only

What does Public Tenders Feed do?

Public Tenders Feed returns new public tenders from four official sources in one record shape: the EU's TED (Tenders Electronic Daily), the UK's Find a Tender, the US SAM.gov Contract Opportunities extract and CanadaBuys. You set the sources, the notice types, your codes (CPV, NAICS, UNSPSC), keywords and buyer countries, run it on a schedule, and each run returns only the notices it has not delivered to you before.

• TED (ted): tenders of buyers in the EU and the EEA, filtered by CPV code, keyword and buyer country.
• Find a Tender (uk_fts): UK tenders above the thresholds, filtered by CPV code and keyword.
• SAM.gov (sam): US federal opportunities from the public extract, filtered by NAICS code and keyword.
• CanadaBuys (canadabuys): Canadian federal tenders, filtered by UNSPSC code and keyword.
• Contracts Finder (uk_cf): the UK's older service for lower-value notices, off unless you add it.

Each row is one notice: its ID and page, the type (open tender, planning, award or update), the title, the buyer and its country, the codes, the value when one is stated, the deadline, the publication date, the place of performance, the procedure, the set-aside for US notices, links to the documents, and which of your codes and keywords it matched. Each source also gets one free summary row per run: what it read, what matched, what was delivered and why the rest was not. A source that fails says why in its summary row, and the other sources still deliver.

No contact names, emails or phone numbers are taken from any source, and descriptions are searched but never copied into the rows. This Actor is not affiliated with or endorsed by the Publications Office of the EU, the UK Cabinet Office, the US General Services Administration or Public Services and Procurement Canada.

Why use Public Tenders Feed?

Finding tenders to bid on means checking four portals with four search forms, four code lists and four date rules, and working out which notices you saw last week. This Actor asks all four for you, puts the notices in one table, remembers what it delivered, and on a schedule sends only what is new.

What you get:

• Four official sources, one record shape. TED, Find a Tender, SAM.gov and CanadaBuys map to the same fields, so one sheet or one alert handles them all.
• Every filter works on every source it applies to. Codes are matched per scheme (CPV, NAICS, UNSPSC), keywords on every source, and each filter has a test per source.
• Only what is new. The memory keeps the notice IDs it delivered, so a notice reaches you once, and a later run reads only from where the last complete run stopped.
• A free summary row per source. It shows how many notices the source had, how many matched and why the others were left out, so a run that finds nothing is visible and costs nothing.
• A failing source does not stop the run. Its summary row says what went wrong, and the next run reads its days again.
• No personal data. No contact details from any source, and no description text.
• No key needed. All four sources are public.

Each claim above has a test behind it:

What you get	How it works	The test that backs it
Four sources, one shape	Each source has its own reader that maps its answer into the same fields. The tests run every reader on real answers saved from the sources on 15 Oct 2026.	"TED: every saved notice maps, with the English title only, the TED page and the PDF link", "Find a Tender and Contracts Finder: release tags give the type, and every saved release maps without contact details", "SAM.gov: every saved row maps; the key holds the type; values only on awards; no contact column is read", "CanadaBuys: every saved row maps; an amended tender also gives an update notice; RFI is planning, ACAN is an award", and the live test "live: the prefill run returns rows and summary rows, and a second run on the same memory repeats nothing".
Filters per source	CPV, NAICS and UNSPSC codes match by prefix, keywords as whole words, types, countries and values the same way everywhere.	"CPV codes (TED, Find a Tender, Contracts Finder) match by prefix, and trailing zeros do not matter", "NAICS codes match SAM.gov notices by prefix", "UNSPSC codes match CanadaBuys notices by prefix, and trailing 00 pairs do not matter", "keywords in titles: every source delivers only notices whose title holds a keyword as a whole word", "excludeKeywords drop notices at every source", "match all needs a code and a keyword; match any takes either", "minValue drops notices with a lower stated value at every source that states values, and keeps the ones without".
Only what is new	The memory holds the IDs delivered in the last 60 days. The window starts where the last complete read of the source started.	"later runs on the same memory never repeat a notice, and the SAM.gov extract is not downloaded again once read in full", "a complete read moves the window: the next run asks Find a Tender only from the last run on", "the window: firstRunDays on a first run, the last complete run later (a day earlier for date-only sources), or since".
The SAM.gov extract read with care	The 200 MB file is streamed row by row. It lists the newest notices first, so reading stops once 1,000 rows in a row are older than the window. An unchanged file is not downloaded again.	"a newest-first extract stops downloading after 1,000 rows older than the window, with every notice of the window read", "a row out of order switches the early stop off, so the whole file is read", "an unchanged extract after a complete read with the same filters is not downloaded", "a SAM.gov download that breaks off is a failed read, not a short file taken as complete".
A failing source does not stop the run	Each source is read on its own. Its summary row carries the reason and its window is kept.	"a source that keeps failing does not fail the run: its summary row says why, the others deliver", "a failed source keeps its window: the next run reads the same days again", "a source that never answers times out, and the others are not held up", "an answer that is not the expected shape fails that source only".
Polite to the sources	TED is paced under 500 notices per 6 minutes and stops at a limit answer. Find a Tender waits as long as its Retry-After header asks. Every request names the Actor.	"TED: reading keeps under 500 notices per 6 minutes (the fair-use limit is 600)", "TED: a limit answer (HTTP 429) stops TED at once, keeps what was read, and says so", "Find a Tender: Retry-After on a 429 is honoured, then reading goes on", "every request names the Actor in its User-Agent".
No personal data	Contact columns and contact points are never read, and descriptions are used for matching only.	"no contact detail from any source reaches a row, the summary or the memory".
Charged only for what you get	One charge per row with status ok. Summary rows are free, and the run stops before it would charge past your maximum cost per run. A run that restarts reads its own rows first.	"the prefill run delivers notices plus a free summary row per source, and charges one event per notice", "your maximum cost per run stops delivery; the next run delivers the rest without repeats", "a restart after a row was written but before the memory was saved charges nothing twice".

Plus what the Apify platform adds around it:

• API access. Start runs and read results with the Apify API or the JavaScript and Python clients.
• Scheduling. Run the feed every weekday morning and get only what was published since the last run.
• Integrations. Send rows to Make, Zapier, Google Sheets, Slack or any webhook.
• Monitoring. Run history, logs and alerts for failed runs.

Typical uses are a daily list of tenders for a bid team, an alert on the codes a company sells in, a feed of awards to see who wins what in a market, and a tender feed for an AI agent that should read only what is new.

What data can Public Tenders Feed extract?

Each notice becomes one row in the dataset. Each source also gets one free summary row per run. The fields:

Field	Type	What it holds
status	string	ok (a notice, charged) or summary (the free row of one source).
source	string	ted, uk_fts, sam, canadabuys or uk_cf.
noticeId, url	string	The source's own ID for the notice and its public page. TED gives a publication number such as 718613-2026, Find a Tender a notice number such as 097467-2026, SAM.gov a 32-character notice ID, CanadaBuys a reference number (with # and the amendment number on an update notice).
noticeType, sourceType	string	open, planning, award or update, and the type in the source's own words, such as competition / cn-standard for TED or Combined Synopsis/Solicitation for SAM.gov.
title	string	The title. For TED it is TED's English heading: the buyer's country, the main CPV label and the buyer's own title, which is often in the buyer's language.
buyerName, buyerCountry	string	The buying organisation, and its country as a two-letter code.
classification	object	scheme (CPV, NAICS or UNSPSC) and codes, the codes of the notice and its lots.
psc	string	SAM.gov only: the product or service code.
value	object	amount and currency when the notice states a value above 0, in the notice's own currency. SAM.gov gives values on awards only, CanadaBuys gives none.
deadline, publishedAt	string	The deadline for bids or responses and the publication date, as the source writes them. TED gives the earliest lot deadline with its time.
placeOfPerformance, procedure, setAside	string	Where the work is done (NUTS codes for TED, regions for the UK and Canada, city and state for the US), the procedure type, and the US small-business set-aside.
documentLinks	array	The notice page, the PDF or the buyer's tender documents, and attachments, at most 20.
matchedKeywords, matchedCodes	array	Which of your keywords and codes the notice matched.
firstSeenAt, licence	string	When this memory first delivered the notice, and the licence line of the source, on every row.
sourceStatus	string	Summary rows: ok (read and delivered in full), partial (read or delivered in part; the rest comes next run), unchanged (the SAM.gov extract has not changed since the last complete read, so it was not downloaded), skipped or failed.
windowFrom, windowTo	string	Summary rows: the time span the source was asked for.
noticesRead, noticesMatched, noticesNew, delivered, alreadySeen, notDelivered	integer	Summary rows: what the source sent, what matched your filters, what was new to the memory, what this run delivered, what was delivered before, and what is left for the next run.
byType, filteredOut	object	Summary rows: the matches per notice type, and the notices left out by reason: window, noticeType, country, value, topic (outside your codes and keywords), excluded (an exclude keyword) and noDate.
filtersApplied, sourceTotal, fileLastModified, durationMs	array, integer, string, integer	Summary rows: your filters as this source used them, TED's own count for the query (or the rows in the SAM.gov and CanadaBuys files read), the extract's Last-Modified for SAM.gov, and the read time.
stateStore, checkedAt	string	The memory's store name, and when the row was written.
message, errorCode	string	Summary rows: what happened in plain words, and a short code when there is something to report.

The Notices view shows the source, type, title, buyer, country, deadline, publication date, value, what matched and the link. The Source summaries view shows each source's counts, status and message.

A summary row that has something to report carries one of these codes in errorCode, so you can count by reason without reading the sentence. The sentence in message can be reworded in a later version. The code stays.

errorCode	Meaning
excluded-by-countries	The source covers one country only, and your countries leave it out, so it was skipped without a request.
no-filter-for-source	You gave codes for other schemes only and no keyword, so this source had nothing to match and was skipped.
max-notices	The run reached maxNotices. The next run delivers the rest.
charge-limit	The run reached your maximum cost per run. The next run delivers the rest.
run-time-limit	The run reached its time limit. The next run reads the window again.
aborted	The run was stopped.
read-limit	The source had more than one run reads (2,000 TED notices, 400 pages from the UK services), or TED's pace would pass the time limit. The next run goes on.
rate-limited	The source answered "too many requests": at once for TED, three times in a row for the others.
server-error, http-error	The source answered with an error status, after three retries for a server error.
network-error, timeout	The source could not be reached, or did not answer or send data in time.
too-large, bad-response, unexpected-redirect	A file over the size limit, an answer not in the expected shape (the source may have changed), or a redirect to another site.

How to get new public tenders by code, keyword and country

1. Open the Actor in Apify Console and select the Input tab.
2. The form is prefilled with a small run: TED and SAM.gov, open tenders, CPV 72 (IT services) for TED, NAICS 5415 (computer systems design) for SAM.gov and the keyword "software" in titles, 3 days back, at most 20 notices. Press Start to see a result straight away. On this computer on 16 Oct 2026 it delivered 20 notices (10 from each source) in about 12 seconds, and the same input run again delivered 20 others and nothing twice.
3. Pick the Sources (sources) and Notice types (noticeTypes): open for tenders you can still bid on, planning for prior information and market research, award for awards, update for changes and cancellations.
4. Add your codes: CPV codes (cpv) for TED and the UK services, NAICS codes (naics) for SAM.gov and UNSPSC codes (unspsc) for CanadaBuys. A short code covers everything under it: CPV 72 is all IT services, 7221 software programming.
5. Add Keywords (keywords), such as cybersecurity or cyber*, and choose Search keywords in (keywordsIn) and Codes and keywords (match). Exclude keywords (excludeKeywords) drop notices that hold them.
6. Optional: Buyer countries (countries), such as DE, FR and IE for TED, and Minimum value (minValue).
7. Set First run days (firstRunDays) and Most notices per run (maxNotices). The first run reads that many days back; later runs read from where the last one stopped (since set to lastRun).
8. Save the input as a task, or set Memory name (stateStoreName), such as my-it-tenders. Then you can change the filters and keep the memory.
9. Press Start. Open the Output tab for the table and export it as JSON, CSV or Excel. Then create a schedule with the same input, for example every weekday at 7:00.

To run it from code, send the same input to the Actor through the API:

{
    "sources": ["ted", "uk_fts", "sam", "canadabuys"],
    "noticeTypes": ["open"],
    "cpv": ["72", "48"],
    "naics": ["5415"],
    "unspsc": ["43"],
    "keywords": ["cyber*", "cloud"],
    "excludeKeywords": ["training"],
    "countries": ["DE", "FR", "GB", "US", "CA"],
    "firstRunDays": 7,
    "maxNotices": 500,
    "stateStoreName": "my-it-tenders"
}

How to send new tenders to Slack or Google Sheets (Make, Zapier, no code)

After the first run, a scheduled feed holds only the notices that are new since the last run, so you can pass every run's results straight on.

Slack, built in. On the Actor's Integrations tab, add Slack, connect your workspace, pick a channel and choose the event "Run succeeded". Each finished run posts a message with a link to its results.

Google Sheets with Make.com.

1. In Make, create a scenario and add the Apify module that watches for finished Actor runs, set to this Actor.
2. Add the Apify module that gets dataset items, and map the run's default dataset ID.
3. Add Google Sheets "Add a Row" and map the fields you want, for example source, title, buyerName, buyerCountry, deadline, value and url.
4. Add a filter so only rows with status ok reach the sheet, or only rows of one noticeType, or only summary rows with an errorCode for an alert.

Google Sheets or Slack with Zapier. Use the Apify trigger for a finished Actor run, then the Apify step that fetches dataset items, then a Google Sheets "Create Spreadsheet Row" or a Slack "Send Channel Message" action. A Zapier filter step works the same way as the Make filter.

Your own code. Add a webhook on the Integrations tab for the event "Run succeeded", and read the dataset from the URL in the payload.

How much does it cost to get new public tenders?

You pay per notice delivered, and the Apify platform usage for the run is included. A notice costs $0.0030 on the Free and Bronze plans, so 1,000 notices cost $3.

Event	When it is charged	Price
notice	One notice row delivered with status ok: each notice of the first run, and each new notice after that, once per memory	$0.0030 (Free, Bronze), $0.0027 (Silver), $0.0025 (Gold)
Actor start	Once per run, scaled by memory (Apify standard event)	$0.00005 per GB of memory

Summary rows cost nothing. You can cap a run with the Maximum cost per run setting, and the Actor stops cleanly at that limit.

How much does one notice cost?

One notice costs $0.0030 on the Free and Bronze plans, $0.0027 on Silver and $0.0025 on Gold. A run adds a start fee of about $0.000025 at the default memory of 512 MB. As a guide: TED listed 72 open tenders for IT services (CPV 72) from German buyers from 12 to 15 Oct 2026, about 18 a publishing day, so about 400 a month: $1.20 a month. The same code across the whole EU gave about 115 a publishing day.

How do I keep a run within budget?

Set Maximum cost per run when you start it or in the schedule, and Most notices per run (maxNotices) in the input. The run stops before it would charge past either, and the sources it did not finish get the code charge-limit or max-notices in their summary rows. Those notices stay out of the memory, so the next run delivers them.

What does a run cost when nothing is new?

Only the start fee, about $0.000025 at 512 MB. Each source writes its free summary row, and nothing is charged.

How often should I run it?

Once a day on weekdays fits the sources: TED publishes its notices once each weekday (they appear the evening before the date they carry), SAM.gov writes its extract once a day, and CanadaBuys writes its list of open tenders once a day. Find a Tender publishes through the day. You pay per notice, not per run, so a run more often costs you only its start fee.

Input

Open the Input tab to see every setting with a short description. No field is required: an empty input reads the four default sources for open tenders from the last run, 7 days back on the first. The form has three parts:

• What to read: sources and noticeTypes.
• What to match: keywords, cpv, naics, unspsc, match, keywordsIn, excludeKeywords, countries and minValue.
• Window and memory: since, firstRunDays, maxNotices and stateStoreName.

sources takes ted, uk_fts, sam, canadabuys and uk_cf. The first four are the default.

noticeTypes takes open (the default), planning, award and update. Each source's own types map to these: for TED, competition notices are open, prior information planning, results and direct award notices award, changes and contract modifications update. For SAM.gov, solicitations and combined synopses are open, presolicitations, sources sought and special notices planning, award notices and justifications award, modifications update. For CanadaBuys, requests for information are planning, advance contract award notices award, an amendment adds an update notice, and every other notice is open.

keywords takes up to 50 words or phrases. A notice matches when one appears as whole words, any case, accents ignored; a phrase matches as neighbouring words, and a word ending in * matches the start of a word. keywordsIn is titleAndDescription (the default) or title. excludeKeywords drop a notice when they appear in the same text.

cpv, naics and unspsc take up to 200 codes each and match by prefix. CPV trailing zeros and its check digit are ignored (72000000-5 works like 72), UNSPSC trailing 00 pairs too. Each source uses its own scheme: CPV for TED, Find a Tender and Contracts Finder, NAICS for SAM.gov, UNSPSC for CanadaBuys.

match is any (the default: one of your codes or one of your keywords) or all (a code of the source's scheme and a keyword). Give no code and no keyword to get every notice of the chosen types. When you give codes for some schemes only and no keyword, a source of another scheme is skipped with no-filter-for-source, so CPV codes alone do not fill your feed with every SAM.gov notice.

countries takes up to 60 buyer country codes, two letters (DE) or three (DEU). TED is asked for those countries. Find a Tender and Contracts Finder (GB), SAM.gov (US) and CanadaBuys (CA) are skipped without a request when their country is not in the list.

minValue drops notices whose stated value is below it, in the notice's own currency. Notices without a stated value are kept, because most open tenders state none.

since is lastRun (the default) or a date such as 2026-10-01. firstRunDays is 7 by default, from 1 to 30. No window reaches more than 30 days back. maxNotices is 500 by default and at most 10,000, all sources together, newest first, the sources taking turns.

stateStoreName names the key-value store that remembers what was delivered. Left empty, a run of a saved task keeps one memory for that task, and a run outside a task keeps one memory per set of filters, so changing a filter there starts a new memory and the next run delivers (and charges) its first run again.

Output

You can download the dataset in various formats such as JSON, HTML, CSV, or Excel. A run also writes a SUMMARY record to its key-value store with the memory's name, the sources read, the notices delivered and charged, the reason the run stopped early if it did, one entry per source, the run time and the peak memory. A run whose sources all failed and that delivered nothing ends with the status Failed and the reason in its message, so a schedule shows it. A wrong input or a wrong billing set-up is refused before any request, in words.

A notice from Find a Tender, the summary of SAM.gov on the first prefill run, and its summary on the next run look like this:

[
    {
        "status": "ok",
        "errorCode": null,
        "message": null,
        "source": "uk_fts",
        "noticeId": "097467-2026",
        "url": "https://www.find-tender.service.gov.uk/Notice/097467-2026",
        "noticeType": "open",
        "sourceType": "tender / UK4",
        "title": "Programme Management Platform",
        "buyerName": "Cheshire West and Chester Council",
        "buyerCountry": "GB",
        "classification": { "scheme": "CPV", "codes": ["72222300"] },
        "psc": null,
        "value": { "amount": 400000, "currency": "GBP" },
        "deadline": "2026-11-12T10:00:00Z",
        "publishedAt": "2026-10-15T13:11:36+01:00",
        "placeOfPerformance": "UKD",
        "procedure": "Open procedure",
        "setAside": null,
        "documentLinks": ["https://www.find-tender.service.gov.uk/Notice/097467-2026"],
        "matchedKeywords": [],
        "matchedCodes": ["72"],
        "firstSeenAt": "2026-10-15T22:45:35.630Z",
        "licence": "Source: Find a Tender (find-tender.service.gov.uk). Contains public sector information licensed under the Open Government Licence v3.0.",
        "sourceStatus": null,
        "windowFrom": null,
        "windowTo": null,
        "noticesRead": null,
        "noticesMatched": null,
        "noticesNew": null,
        "delivered": null,
        "alreadySeen": null,
        "notDelivered": null,
        "byType": null,
        "filteredOut": null,
        "filtersApplied": null,
        "sourceTotal": null,
        "fileLastModified": null,
        "durationMs": null,
        "stateStore": "my-it-tenders",
        "checkedAt": "2026-10-15T22:45:35.630Z"
    },
    {
        "status": "summary",
        "errorCode": null,
        "message": "SAM.gov (US), notices published from 2026-10-12 (first run): 4,588 notices read, 10 matched your filters (10 open), 0 delivered before, 10 new, 10 delivered. Left out: 1,000 published before the window, 1,128 of other notice types, 2,450 outside your codes and keywords. The extract lists the newest notices first, so reading stopped after 1,000 notices older than the window (11.2 MB of 202.6 MB downloaded). Free row.",
        "source": "sam",
        "sourceStatus": "ok",
        "windowFrom": "2026-10-12T22:26:54.963Z",
        "windowTo": "2026-10-15T22:26:54.963Z",
        "noticesRead": 4588,
        "noticesMatched": 10,
        "noticesNew": 10,
        "delivered": 10,
        "alreadySeen": 0,
        "notDelivered": 0,
        "byType": { "open": 10, "planning": 0, "award": 0, "update": 0 },
        "filteredOut": { "window": 1000, "noticeType": 1128, "country": 0, "value": 0, "topic": 2450, "excluded": 0, "noDate": 0 },
        "filtersApplied": ["types: open", "NAICS: 5415", "keywords (title): software", "match: any"],
        "fileLastModified": "Thu, 15 Oct 2026 03:30:31 GMT",
        "durationMs": 12151
    },
    {
        "status": "summary",
        "message": "SAM.gov (US): the extract has not changed since the last complete read (Last-Modified Thu, 15 Oct 2026 03:30:31 GMT), so it was not downloaded: there is nothing new. Free row.",
        "source": "sam",
        "sourceStatus": "unchanged",
        "noticesRead": 0,
        "delivered": 0
    }
]

The rows come from local runs on 16 Oct 2026 against the real sources. The summary rows are shortened. Every row carries every field, with null where a value does not apply, so exports keep the same columns.

Tips for faster runs and lower costs

• Save the input as a task, or set stateStoreName, before you change the filters. Outside a task and without a name, a change starts a new memory, and the next run delivers and charges its first run again.
• For SAM.gov, prefer NAICS codes or keywordsIn title. SAM.gov descriptions are long and carry standard wording, such as a line on the software needed to open the attachments, so a keyword searched in descriptions finds notices on other subjects. On 16 Oct 2026, 5 of the 10 SAM.gov notices that "software" found in descriptions were about snow removal, bearings and equipment.
• Keep the first run small with firstRunDays and maxNotices, then let the schedule bring the rest.
• Use countries to skip whole sources: with ["DE", "FR"] only TED is read.
• A first run of all four sources 7 days back, all notice types, took about 2.5 minutes on this computer on 16 Oct 2026, most of it a 2-minute wait that Find a Tender asked for with its Retry-After header. A daily run of the same codes took about 12 seconds.
• TED is read at most 500 notices per 6 minutes, below its fair-use limit. A catch-up that needs more than that waits, and a run that read 1,000 TED notices took about 6.5 minutes.
• The default 512 MB is plenty: the SAM.gov extract is streamed, and the highest peak measured was 144 MB.

FAQ, disclaimers and support

Where does the data come from, and may I use it?

From the official publishers, through their public interfaces and files:

• TED: the TED search API of the Publications Office of the European Union. Reuse is allowed under Commission Decision 2011/833/EU, with the source named.
• Find a Tender and Contracts Finder: the UK services' OCDS interfaces. Public sector information licensed under the Open Government Licence v3.0.
• SAM.gov: the Contract Opportunities public data extract on data.gov. US federal government data, public domain.
• CanadaBuys: the open data files of tender notices. Information licensed under the Open Government Licence - Canada.

Every row carries its source's licence line in licence. Keep it when you pass the rows on. How you may use what you collect depends on your purpose and your country, so check that for your use case. This is general information, not legal advice.

Is this an official tool of any of the publishers?

No. The Actor is an independent tool, not affiliated with, made, endorsed or checked by the Publications Office of the EU, the UK Cabinet Office, the US General Services Administration or Public Services and Procurement Canada. The data comes as the publishers give it. Open the notice page (the url column) and the tender documents before you bid or act on a row.

What counts as new, and where can it be wrong?

A notice is new when the memory has not delivered it before. The memory knows a notice by the TED publication number, the Find a Tender or Contracts Finder release ID, the SAM.gov notice type and ID, or the CanadaBuys reference number (with the amendment number for an update notice). The memory keeps the keys delivered in the last 60 days, at most 100,000 per source, longer than any window reaches back. Each source reads from the start of its last complete read, a day earlier for the sources that give dates without a time, and the repeats this brings are counted in alreadySeen and not delivered. A read is complete when the source was read to the end and every notice that matched was delivered. Where the result can be wrong:

• One tender, several notices. A tender, its changes and its award are separate notices with their own IDs at every source, so with award and update on you get each of them once.
• A SAM.gov notice that changes type. A solicitation that becomes an award notice keeps its ID, and the type is part of the key, so the award notice is delivered as new.
• A new memory. A changed set of filters outside a task and without stateStoreName starts a new memory, which delivers its first run again.
• A notice older than the window. A notice published before the window, or one a source adds late with an old date, is not read.
• A run that stops early. A source cut by maxNotices, the maximum cost per run or the time limit keeps its window, and the next run delivers what was left.

How does a notice match, and where can it be wrong?

Codes match by prefix in the scheme of the source: CPV for TED and the UK services, NAICS for SAM.gov, UNSPSC for CanadaBuys. A notice's codes are those of the notice and of its lots or items. Keywords match whole words in the title and, with titleAndDescription, in the description: TED's English heading and lot titles and its procedure and lot descriptions, the UK tender title and description, the SAM.gov title and description, the CanadaBuys title and description in English and French. Text is compared in lower case without accents. match any takes a code or a keyword, all needs both. Where the result can be wrong:

• SAM.gov descriptions. They carry standard wording, so a keyword can match a notice on another subject (see the tips).
• TED titles in other languages. TED's English heading names the country and the main CPV label in English, then the buyer's own title, often in the buyer's language. An English keyword matches the label and English titles, not a German title. Codes work across languages.
• Notices without codes. Many Find a Tender award releases list no CPV code, so a CPV filter misses them. Add a keyword to find them.
• Codes the buyer chose. Each buyer picks its own codes, and a notice for IT support may carry only a building services code.
• Values. minValue compares the stated value in the notice's own currency, without conversion. Most open tenders state no value and are kept.
• CanadaBuys types. A CanadaBuys notice without a notice type counts as open. Notices without a publication date are left out and counted in filteredOut as noDate.

How fresh are the notices?

TED's notices carry the date of the next weekday and appear in its search the evening before: a run at 22:21 UTC on 15 Oct 2026 delivered notices dated 16 Oct. Find a Tender publishes through the day, and a run reads its notices up to the time the run starts. SAM.gov writes its extract once a day (the 15 Oct 2026 file was written at 03:30 UTC), so a SAM.gov notice reaches you up to a day after it was posted. CanadaBuys writes its file of open tenders once a day and its file of new tenders through the day, and the Actor reads both. Where the result can be wrong:

• SAM.gov notices that close fast. The extract lists active notices, so a notice posted and closed between two files is never in it.
• Time zones. SAM.gov gives its posting time without a time zone, and CanadaBuys and TED give dates, so the window for those sources starts on a whole day (UTC).

Is personal data involved?

The Actor takes no contact names, emails or phone numbers from any source, and it searches descriptions without copying them, because they often name a contracting officer. Buyer names are organisations. A title or a buyer name is published as the buyer wrote it, and a few can name a person, such as a sole trader. If a row holds personal data it should not, open an issue on the Issues tab with the notice ID, and the field will be fixed in the Actor.

How does the Actor remember what it delivered, and how do I start over?

After each run it keeps, per source, where its next window starts, the SAM.gov extract's Last-Modified, and the notice IDs it delivered with the day of delivery (the last 60 days, at most 100,000 per source). They are kept in a named key-value store listed in each row as stateStore. No notice text is stored there. To start over, delete that store under Storage in Console, or set a new stateStoreName.

What happens when a source fails or limits the requests?

Each source is read on its own, one request at a time, with a User-Agent that names the Actor. A server error or a network error is retried three times, after 2, 4 and 8 seconds. Find a Tender's "too many requests" answer (HTTP 429) is waited out for as long as its Retry-After header says, up to 2 minutes, and three in a row stop it for the run. TED is read at most 500 notices per 6 minutes, below its fair-use limit of 600, and a limit answer stops TED for the run at once. A page that does not answer within 60 seconds, or a download that sends nothing for 90 seconds, is a timeout. In every case the source's summary row says what happened, its window is kept so the next run reads those days again, and the other sources deliver as usual. The run fails only when every source failed and nothing was delivered.

What happens when a run reaches its time limit?

The Actor reads the time limit of the run from the platform and starts no new work in the last 45 seconds. What it did not finish gets the code run-time-limit in its summary row, and the next run reads that window again.

What if something looks wrong?

Open the Issues tab on the Actor page and describe your input and what you expected. A run link helps. Issues are answered within two days.

Other Actors by the same developer

• News Monitor: new news articles for the companies, people, keywords and GDELT themes you watch, from the GDELT Project's 15-minute news files, with only articles not delivered before.
• New Scholarly Papers Watch: new journal articles, preprints and other works for the keywords, ORCID iDs, journals and funders you watch, with retraction and correction notices for your DOIs.
• SEC EDGAR Filing Monitor: new SEC filings for the tickers or CIKs you watch.
• SEC Form D Feed: new private fundraising rounds filed with the SEC on Form D, with amounts and industry.
• SEC XBRL Fundamentals: revenue, EPS, cash flow and other reported figures per company and period, from SEC XBRL filings.
• UK Company Check: Companies House status, outstanding charges and risk flags for UK companies, or a list of newly registered ones.
• US Trademark Status and Watch: USPTO status and upcoming deadlines for US trademarks, and what changed since your last run for marks you watch.
• Full Page Website Screenshot: full-page PNG or JPEG screenshots, including pages that load content as you scroll.