EU Tender Supplier Fit & Requirements Review

Review procurement lots against your company profile before investing time in a bid. Get source conditions, exact quotations, deadline exclusions, relevance components, conflicts with supplied company facts, and explicit unknowns.

What the Actor does

• Searches the official EU TED API, or accepts explicit lot records from your own system or another Actor.
• Reads official XML when available and keeps lot-specific requirements, classifications, locations and deadlines separate. Procedure-level exclusion declarations are labelled separately.
• Excludes award/result notices, passed submission deadlines and explicitly conflicting place-of-performance countries.
• Calculates a transparent relevance score: 40 for an exact supplied CPV match, 40 for a supplied service phrase found in this lot's title/description, 20 for a supplied performance-country match. Empty or missing values earn no points. This score does not measure legal eligibility or probability of winning.
• Optionally uses Anthropic to compare multilingual source conditions with your supplied company facts. Every accepted source and company quotation is checked against its actual input text.
• Supports snapshots and changes, including deadline expiry without edits to the source.

Every non-excluded result remains review-required. legalEligibilityConfirmed and documentsComplete are always false. A company fact is a self-reported declaration, not an independently verified certificate.

Start without an AI key

{
  "companyProfile": {
    "name": "Example security integrator",
    "services": ["camera installation", "access control"],
    "cpvCodes": ["35120000", "32323500"],
    "countries": [],
    "facts": [
      { "id": "services", "text": "We install and maintain video surveillance systems." }
    ]
  },
  "maxNotices": 5,
  "maxXmlDocuments": 3,
  "maxLots": 20
}

Replace the example profile with your company's facts. Search defaults to the last thirty days and selected security-related CPV codes. Set cpvCodes to your desired search categories or an empty array for no category restriction. Search categories and profile relevance categories are separate. noticeIds selects exact TED publication numbers and replaces date/category discovery.

companyProfile.countries restricts place of performance, using three-letter country codes. It does not claim that a company is eligible based on its nationality. Buyer countries are returned separately.

Anthropic semantic review

Set analysisMode to anthropic, enter your own encrypted anthropicApiKey, choose a compatible model, and supply explicit companyProfile.facts with unique IDs. All generated titles and reasoning are English; original source quotations retain their source language.

Requirement statuses:

Status	Meaning
profile-supported	A supplied company fact supports the condition, with an exact fact quotation. Not independently verified eligibility.
conflict	An explicit supplied fact conflicts with the source condition. Review the evidence.
unknown	Missing, ambiguous, unreviewed or unverifiable evidence. Missing facts do not become negative facts.

The model receives up to ten conditions per lot, up to 2500 characters per condition, and at most thirty supplied facts. Selection conditions and lot-specific conditions are prioritized, with lexical relevance to the supplied facts as a tie-breaker. The full returned source conditions remain available; unreviewed conditions stay unknown. The Actor validates identity and literal quotations, not the correctness of every semantic or legal interpretation. Source instructions are treated as data. Review important decisions yourself.

maxAiLots and maxAiBudgetUsd bound requests using the token prices you configure. The cost estimate uses a conservative pre-request allowance, accounts for rejected/incomplete replies, and reserves the allowance when usage is unknown. It is not a hard Anthropic account spending cap. Set correct prices and your provider's own spending controls. There are no automatic provider retries.

Use another Actor's data

Set sourceMode to provided-notices. Each notice contains id, sourceUrl, noticeType and an explicit lots array. A lot contains:

{
  "id": "LOT-0001",
  "titleOriginal": "Camera installation",
  "descriptionOriginal": "Installation and maintenance of cameras.",
  "cpvCodes": ["35120000"],
  "countryCodes": ["ROU"],
  "deadlineDateOriginal": "2030-12-01+02:00",
  "deadlineTimeOriginal": "15:00:00+02:00",
  "requirements": [
    { "id": "authorization", "textOriginal": "The bidder must hold an ANRE C1A authorization." }
  ]
}

This is an illustration, not a real tender. See example-provided-input.json for a complete sample. The Actor does not fetch arbitrary provided URLs or launch paid child Actors. The caller obtains data with appropriate permissions and supplies the lot structure. A flattened feed of parallel arrays cannot establish lot-to-deadline or lot-to-requirement relationships.

Coverage and missing XML

TED search is a bounded window, not exhaustive coverage of all procurement portals. XML may be unavailable. With exactly one explicit lot identifier, the Actor can retain that lot's API description, source criteria and a single published date, labelled single-lot-api. Exact time, attachments and criteria completeness remain unverified. With multiple or missing lot identifiers it returns an unverified scope, unknown deadline and no inferred detailed requirements. It never joins parallel arrays by position. Missing fields do not mean no legal conditions exist. Procurement attachments, documents on other portals, amendments, cancelled procedures and company documents require separate verification. This version does not download PDFs or promise full UK coverage. API fallbacks do not replace a verified XML baseline in change mode.

Limits: up to fifty source notices, ten search pages, ten XML downloads and one hundred assessed lots. The summary reports search-window coverage, XML attempts, delivered rows and AI usage. Source failures do not produce a false successful full-source baseline.

Changes and state

Use stateStoreId for a dedicated workflow. First run with mode: "snapshot", then use mode: "changes". Only delivered, sufficiently scoped and successfully reviewed rows advance state. Unverified XML fallback or failed/incomplete AI reviews remain retryable. Billing limits cannot mark undelivered records as delivered. Company profile and model changes use a different state namespace.

Run the same state serially. A pre-write check detects observed concurrent changes, but the storage API does not provide an atomic distributed lock. Do not overlap runs sharing a store. A failure after dataset delivery but before state persistence can cause redelivery; downstream consumers should deduplicate by id and sourceHash. No deletion or withdrawal inference is made from absence in a bounded search.

Pricing

Proposed initial price: $0.01 per start + $0.02 per delivered lot review. Excluded lots and explicitly unverified-scope results are assessments and are charged when delivered. A run that delivers zero rows still incurs the start fee. Platform usage is included in this proposed Actor price. Pricing is active only when configured in Apify; check the Store pricing panel for current terms.

Examples excluding external AI: zero rows $0.01; ten rows $0.21; one hundred rows $2.01; one thousand rows across multiple bounded runs incur each run's start fee.

Anthropic usage is optional, paid directly through your own key, and is additional to Actor charges. This Actor does not purchase upstream Actor outputs for you.

Outputs

The default dataset contains one assessment per lot or one explicitly unverified fallback scope. SUMMARY in the default key-value store describes bounds, source coverage, delivered count and estimated AI cost. Original titles and quotations preserve source language. All keys, generated titles, explanations and runtime messages are English.

This is preliminary decision support, not legal advice, procurement eligibility certification, an assurance of completeness, or a guarantee of winning a contract. Always verify the official notice and procurement documents.

Sources: TED Search API, TED legal notice. Rights in documents on third-party portals must be checked separately.