Wikidata Scraper & Entity Resolver — Q-ID, Facts & Cross-Database IDs

A Wikidata scraper and entity-resolution tool built for AI agents, LLMs and RAG pipelines. Resolve any person, company, place or concept to a single structured Wikidata entity — its Q-ID, label, description, type, key facts and cross-database identifiers (VIAF, ISNI, ORCID, IMDb, socials, GND and more). It is exactly the "who/what is X?" lookup an agent needs to disambiguate, ground and enrich entities — no login, no coding, no API key. This Wikidata scraper turns a fuzzy name into a stable, sourced, machine-readable record.

Export to JSON, CSV, Excel, or API — or call it live as an MCP tool.

Built for AI engineers, knowledge-graph teams, data journalists, OSINT & due-diligence analysts, librarians, and anyone enriching records with canonical entity data. If you need to scrape Wikidata at scale or answer "which Elon Musk? what is Apple's founder, industry, VIAF id? what type of thing is Berlin?", this is your grounding store.

---

🔎 Why scrape Wikidata?

Wikidata is the free, collaborative knowledge base behind Wikipedia — a structured graph of over 100 million entities (people, organizations, places, works, species, concepts) with stable identifiers and machine-readable facts. Every entity has a permanent Q-ID (like Q317521 for Elon Musk), a label and description in hundreds of languages, typed statements (date of birth, country, industry, founder, population…), and links out to dozens of external authority databases (VIAF, ISNI, ORCID, IMDb, LinkedIn, MusicBrainz, GND, Library of Congress…).

That makes Wikidata the single best free grounding source for entity resolution: given a messy name, you get back one canonical ID plus facts you can trust and cross-reference. The problem is that the raw Wikidata API speaks in cryptic property codes (P31, P106, P569) and nested JSON — resolving instanceOf → human or occupation → [programmer, engineer] by hand is slow and error-prone. Doing it for a whole spreadsheet of names, in the right language, with the right identifiers pulled out, is a real engineering job.

This Wikidata scraper does that work for you. It searches, disambiguates, fetches, and flattens every entity into a clean record — labels instead of Q-IDs, identifiers named instead of numbered — ready to drop into an LLM context window, a knowledge graph, or a database column.

Common reasons people scrape Wikidata:

• 🤖 Ground an AI agent / LLM — turn an extracted name into a stable Q-ID + facts so the model stops hallucinating identities.
• 🔗 Entity linking & record matching — map your rows to VIAF, ISNI, ORCID, IMDb, GND and other authority IDs.
• 🧩 Build or enrich a knowledge graph — pull typed facts (type, country, industry, founders, inception) for thousands of entities.
• 📰 Fact-check & research — verify who someone is, what a company does, or what category a thing belongs to, with a sourced link.
• 🕵️ OSINT, KYC & due diligence — resolve a name to official website, social handles and external registries.
• 📚 Library & archive cataloging — reconcile names against Wikidata and pull authority-file identifiers.
• 🌍 Multilingual lookups — get the same entity's label and description in German, French, Spanish, Turkish, Japanese and more.

---

✨ What this Wikidata scraper does

Give it a list of names (or specific Q-IDs) and, for each one, it returns a clean, agent-friendly entity record:

• ✅ Name → stable Q-ID — it searches Wikidata and picks the best-matching entity, the disambiguation an LLM can't do reliably alone.
• ✅ Label & description — the canonical name and a short human description, in your chosen language (falls back to English when a language is missing).
• ✅ Type — what the entity is (human, enterprise, city, mountain, chemical process…), resolved to a readable label.
• ✅ Aliases — also-known-as names for better matching and display.
• ✅ Key facts — occupation, birth/death dates, citizenship, birthplace, country, headquarters, founders, industry, legal form, inception, capital, population, author, director, performer, creator, manufacturer — resolved to readable labels, not raw Q-IDs.
• ✅ Cross-database identifiers — VIAF, ISNI, ORCID, IMDb, Twitter/X, Facebook, Instagram, YouTube, LinkedIn, TikTok, GND, Library of Congress, MusicBrainz, Spotify, Discogs, Quora, GRID, Ringgold — whichever the entity has.
• ✅ Official website + Wikipedia + Wikidata URLs — the source links for every record.
• ✅ Direct Q-ID mode — skip search and resolve exact entities when you already know the Q-ID.
• ✅ Multilingual — choose from 20 languages for labels, descriptions and facts.
• ✅ Live MCP / Standby API — besides classic batch runs, the actor can stay warm as an HTTP API and answer entity lookups instantly (ideal for agent tool-calls).
• ✅ Fast, keyless & reliable — no API key, no anti-bot, no proxy required; built-in retries with backoff; output ready as JSON, CSV, Excel or API.

Every claim here is backed by what the actor actually captures — no promised fields it doesn't return.

---

🎯 Use cases

• AI agents & copilots — a tool-call that resolves any entity mention to a grounded Q-ID + facts before answering.
• RAG pipelines — attach canonical entity metadata to retrieved chunks to reduce ambiguity and hallucination.
• Knowledge-graph construction — bulk-resolve a list of entities into typed nodes with external-ID edges.
• Data journalism & fact-checking — confirm identities, affiliations and categories with a sourced Wikidata link.
• KYC / AML / due diligence — resolve names to official websites, social handles and authority identifiers (pairs well with sanctions & LEI lookups).
• Library science & digital humanities — reconcile author/person names against VIAF, ISNI, ORCID and GND.
• Market & company research — pull industry, founders, headquarters, legal form and inception for a list of firms.
• Dataset enrichment — add a qid, type, description and identifier columns to any spreadsheet of names.

---

📥 Input

Provide a list of names to resolve (or exact Q-IDs), pick a language, and optionally cap the result count. Everything is optional — an empty run resolves a ready-made sample of well-known entities so you always get useful output.

Field	Type	Description
queries	array	Names to resolve, e.g. ["Elon Musk", "Apple Inc", "Berlin"]. Each name is matched to the best Wikidata entity. Leave empty for a diverse sample.
qids	array	Optional (advanced): resolve specific Wikidata Q-IDs directly, e.g. ["Q317521", "Q95"]. Bypasses search.
language	string	Label/description/fact language — one of 20 options (en, de, fr, es, it, pt, nl, pl, ru, tr, ar, ja, zh, ko, hi, sv, uk, id, vi, fa). English is the most complete. Default en.
includeFacts	boolean	Add readable key facts (type, occupation, dates, country, industry, founders, website…). Default true.
maxResults	integer	Maximum number of entities to return. Set 0 for no limit. Default 0.
proxyConfiguration	object	Optional Apify Proxy. Off by default — Wikidata is keyless with no anti-bot.

Example 1 — Resolve several names

{
  "queries": ["Elon Musk", "Apple Inc", "Berlin"],
  "language": "en",
  "includeFacts": true,
  "maxResults": 100
}

Example 2 — Direct Q-IDs, German labels

{
  "qids": ["Q317521", "Q312"],
  "language": "de",
  "includeFacts": true
}

---

📤 Output

Each resolved entity is one clean record. The facts and externalIds fields are returned as JSON strings so they stay compact in a table yet carry the full structured detail when parsed.

{
  "query": "Elon Musk",
  "qid": "Q317521",
  "label": "Elon Musk",
  "description": "American businessman (born 1971)",
  "type": "human",
  "aliases": "Elon Reeve Musk",
  "officialWebsite": null,
  "wikipediaUrl": "https://en.wikipedia.org/wiki/Elon_Musk",
  "wikidataUrl": "https://www.wikidata.org/wiki/Q317521",
  "facts": "{\"instanceOf\":\"human\",\"occupation\":[\"programmer\",\"engineer\"],\"birthDate\":\"1971-06-28\",\"citizenship\":[\"South Africa\",\"Canada\",\"United States\"]}",
  "externalIds": "{\"viaf\":\"306339060\",\"orcid\":\"0000-0001-8013-3548\",\"imdb\":\"nm1907769\",\"twitter\":\"elonmusk\",\"gnd\":\"1062481119\"}",
  "scrapedAt": "2026-10-25T00:00:00.000Z"
}

Output fields

Field	Description
query	The name or Q-ID that was resolved (your input echoed back).
qid	Wikidata entity Q-ID (stable permanent identifier).
label	Canonical entity label in the chosen language (English fallback).
description	Short Wikidata description of the entity.
type	What the entity is (instance-of), as a readable label — e.g. human, city.
aliases	Also-known-as names, comma-separated (present when available).
officialWebsite	Official website URL (present when the entity has one).
wikipediaUrl	Link to the matching Wikipedia article (present when a sitelink exists).
wikidataUrl	Link to the entity's Wikidata page.
facts	Key facts as a JSON string — type, occupation, dates, citizenship, country, headquarters, founders, industry, inception, population and more (present when includeFacts is on and the data exists).
externalIds	Cross-database identifiers as a JSON string — VIAF, ISNI, ORCID, IMDb, socials, GND, MusicBrainz, etc. (present when the entity has them).
scrapedAt	ISO timestamp of when the record was resolved.

Fields like officialWebsite, aliases, externalIds and individual facts are data-dependent — they appear only when Wikidata actually holds that information for the entity (a chemical process has no occupation; a private person may have no social handles). Core fields (query, qid, label, type, wikidataUrl, scrapedAt) are present for every resolved entity.

---

🔁 Pagination & scale

This is an entity resolver, not a page-crawler: you hand it a list of names and/or Q-IDs and it resolves every one of them in a single run. There are no result "pages" to walk — the batch is your input list.

Under the hood the scraper is efficient at scale: names are searched in parallel, entity data is fetched in batches of up to 50 Q-IDs per API call, and all referenced fact-entities (for readable labels) are resolved in one more batched pass. Supply thousands of names and it works straight through them. Use maxResults to cap how many entities are returned (0 = no limit), and output is written to the dataset where you can export JSON, CSV, Excel or pull it via the Apify API. Built-in retries with exponential backoff (and optional proxy rotation on HTTP 429) keep large runs reliable.

---

⚙️ How to use

1. Open the actor on the Apify Console and paste your names into Names to resolve (or Q-IDs into the Advanced box).
2. Pick a Language and leave Include key facts on for the richest output.
3. (Optional) Set Max results to cap the run, or leave 0 for everything.
4. Click Start, then read the structured entities from Storage → Dataset.
5. Export as JSON, CSV or Excel, or fetch results through the Apify API / JS & Python clients.
6. Automate it: run on a schedule, trigger via API/webhook, or wire it into Make, Zapier, n8n or Google Sheets.
7. Call it as an AI tool: every Apify actor is MCP-callable via mcp.apify.com — add this Wikidata resolver as a tool in Claude Desktop, VS Code, LangChain (ApifyActorsTool), LlamaIndex, CrewAI or the Vercel AI SDK.

---

💡 Tips

• Pass Q-IDs for exact control. Search picks the top match; if you already know the entity, use qids to avoid any ambiguity.
• Keep includeFacts on unless you only need IDs and labels — it adds type, occupation, country, founders, industry and more at almost no cost.
• English is the most complete language. Non-English runs fall back to English labels/descriptions when a translation is missing, so you never get blanks for core fields.
• Parse facts / externalIds with JSON.parse (JS) or json.loads (Python) to turn the compact strings into objects for your pipeline.
• Disambiguate tricky names by adding context — "Michael Jordan (basketball)" or just pass the right Q-ID.
• No proxy needed. Wikidata is keyless and anti-bot-free; leave proxy off unless your network specifically requires it.

---

❓ FAQ

Is scraping Wikidata legal? Yes. This actor uses the public, official Wikidata API, and Wikidata's content is released under CC0 (public domain). It sends a descriptive user agent and respects the API's rate limits.

Do I need an API key or login? No. Wikidata is keyless and has no anti-bot wall — there is nothing to authenticate and no proxy is required.

How many entities can I resolve, and how fast? As many as you supply — from a handful to thousands. The scraper searches names in parallel and batches entity fetches (up to 50 per call), so small jobs finish in seconds and large lists run straight through. Use maxResults to cap output.

Do I need proxies? No. The resolver runs fine without any proxy. A proxy option exists for unusual network setups, but it's off by default.

What export formats are supported? JSON, CSV, Excel (XLSX), HTML and the Apify API. You can also fetch results with the JS or Python Apify clients.

Is there an API, and can an AI agent call it directly? Yes. Run it via the Apify API, and because every Apify actor is MCP-callable (mcp.apify.com) you can register this Wikidata resolver as a tool for Claude, ChatGPT, LangChain, CrewAI, LlamaIndex or any MCP client. It also supports a warm Standby HTTP mode for instant, low-latency lookups.

Are the facts human-readable? Yes. Referenced entities (occupation, country, founder, industry…) are resolved to labels, not raw Q-IDs — so you get "occupation": ["engineer"], not "P106": ["Q81096"].

Which cross-database identifiers can it return? VIAF, ISNI, ORCID, IMDb, Twitter/X, Facebook, Instagram, YouTube, LinkedIn, TikTok, GND, Library of Congress, MusicBrainz, Spotify, Discogs, Quora, GRID and Ringgold — whichever the entity actually has.

How fresh is the data? It reads live from Wikidata on every run, so results reflect the current state of the knowledge base, with a scrapedAt timestamp on each record.

Can I get labels in another language? Yes — choose from 20 languages (German, French, Spanish, Italian, Portuguese, Dutch, Polish, Russian, Turkish, Arabic, Japanese, Chinese, Korean, Hindi, Swedish, Ukrainian, Indonesian, Vietnamese, Persian and English). Missing translations fall back to English.

What happens if a name has no match? That name is skipped with a warning and the run continues — you still get clean records for every name that did resolve.

---

🔗 Related scrapers

Build richer entity, company and OSINT pipelines by combining this Wikidata resolver with:

• SEC EDGAR Scraper — US Public Company Filings & Profiles
• GLEIF LEI Lookup & Corporate Ownership for AI Agents
• AML & Sanctions Screening for AI Agents (OFAC, EU, UN, UK)
• Genius Scraper — Lyrics, Credits, Samples & Annotations
• Flight Tracker — Live Aircraft Data for AI Agents
• Baltic & Nordic Vessel Tracker — Live AIS for AI Agents

---

Resolve any name to a canonical Wikidata entity — Q-ID, facts and cross-database IDs — with no key and no code. ▶️ Click Start and ground your data. 🧠