Wikidata Entity Resolver for AI Agents

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

A high-speed Wikidata entity resolver and knowledge-graph extractor built for AI agents, LLMs, and enterprise RAG pipelines. Disambiguate and resolve any person, company, geographical place, or conceptual topic to a canonical Wikidata entity — extracting permanent Q-IDs, labels, descriptions, typed classes, key facts, and cross-database identifiers (VIAF, ISNI, ORCID, IMDb, LinkedIn, Twitter/X, and GND).

Whether resolving messy strings for master data management, grounding agentic workflows via Model Context Protocol (MCP), or batch-enriching 1,000+ entities, everything outputs cleanly into JSON, CSV, Excel, or programmatic API.

---

⚡ At a glance

Feature	Specification
Knowledge Base	Official Wikidata API (www.wikidata.org/w/api.php)
Data Scope	Q-ID, canonical label, description, type, aliases, official website, Wikipedia link, facts, cross-IDs
Authentication	Keyless (no API key, account, or billing credentials required)
Entity Types	People, corporations, startups, cities, countries, landmarks, creative works, scientific concepts
Fact Label Resolution	Automatically resolves internal Q-ID claims into human-readable strings (e.g. P106: Q81096 → engineer)
External Authority IDs	VIAF, ISNI, ORCID, IMDb, Twitter/X, Facebook, Instagram, YouTube, LinkedIn, MusicBrainz, GND
Multilingual Support	20 languages supported (English, German, French, Spanish, Japanese, Chinese, Turkish, etc.)
Batch Capacity	Fully unlocked for 1,000+ entities per run with bounded-concurrency pooling
Execution Modes	Classic asynchronous batch jobs or warm low-latency HTTP Standby API
Export Formats	Dataset JSON, flat CSV, spreadsheet Excel, XML, HTML, Webhook integrations

---

🔎 Why resolve entities against Wikidata?

Wikidata represents the single most comprehensive open knowledge graph on earth, containing over 110 million interconnected entities. Unlike unstructured web pages, Wikidata provides persistent, language-agnostic identifiers (Q-IDs) paired with rich semantic statements.

However, querying raw Wikidata endpoints programmatically presents major friction for developers:

1. Cryptic property codes: Raw statements are encoded as abstract alphanumeric keys (e.g., P31 for instance-of, P106 for occupation, P159 for headquarters).
2. Unresolved claim targets: Claim values often reference other Q-IDs rather than human-readable text, requiring recursive batch lookups.
3. Throttled burst requests: Firing hundreds of concurrent search queries against wbsearchentities triggers aggressive HTTP 429 rate limits and dropped connections.
4. LLM hallucinations: Without an external grounding anchor, generative models frequently hallucinate biographical details, executive titles, or corporate ownership.

This Wikidata entity resolver eliminates every layer of complexity. It handles search disambiguation, batch claim fetching (up to 50 entities per network call), and recursive label resolution behind the scenes, delivering flat, analytics-ready records ready for your data pipelines or agent context windows.

---

⚖️ Feature comparison: Why choose this Actor?

Capability	Raw Wikidata API	Manual Wikipedia Search	This Apify Actor
Identifier Disambiguation	Returns raw Q-IDs	Manual copy-paste	Automatic string → Q-ID matching
Fact Humanization	Cryptic P31, P106 codes	Reading freeform text	Readable labels (e.g. "software company")
Cross-Database IDs	Scattered across claims	Incomplete	Unified JSON (VIAF, ISNI, ORCID, IMDb)
Batch Scalability	Prone to 429 rate limits	Extremely slow	1,000+ entities via bounded pooling
AI Agent Readiness	Heavy preprocessing needed	Unusable	Native MCP tool & RAG-ready fields
Export Formats	Nested raw JSON only	None	One-click CSV, Excel, JSON, API

---

✨ Core capabilities

• 🎯 Fuzzy name disambiguation: Matches messy input strings (e.g., "Apple", "Elon", "Paris") to their highest-relevance canonical Wikidata Q-IDs.
• 🏷️ Human-readable fact synthesis: Automatically translates raw property-value tuples into plain language (e.g., inception: "1976-04-01", founders: ["Steve Jobs", "Steve Wozniak"]).
• 🔗 Comprehensive authority linking: Gathers cross-database identifiers including VIAF, ISNI, ORCID, IMDb, GND, and verified corporate social profiles.
• 🌐 True multilingual extraction: Request labels, descriptions, and facts in any of 20 major languages with seamless English fallbacks.
• 📦 1,000+ entity batching: Optimized request batching handles hundreds or thousands of seed names with zero rate-limit drops.
• 🤖 Agentic & MCP tool calling: Plug this actor directly into Claude Desktop, Cursor, LangChain, or CrewAI via Apify's Model Context Protocol (MCP) bridge.
• ⚡ Zero-proxy efficiency: Connects directly to Wikidata's public infrastructure with zero proxy costs and sub-second average response times.

---

🏛️ Supported facts & authority mappings

The resolver parses both semantic knowledge facts and external authority identifiers, transforming Wikidata's cryptic P-codes into intuitive JSON keys:

🧠 Semantic knowledge facts (facts object)

Property Code	Output Key	Description	Example Value
P31	instanceOf	Entity class / type	"technology company", "human"
P106	occupation	Professional role	["physicist", "chemist"]
P569	birthDate	Date of birth (YYYY-MM-DD)	"1867-11-07"
P570	deathDate	Date of death (YYYY-MM-DD)	"1934-07-04"
P27	citizenship	Country of legal citizenship	["Poland", "France"]
P19	birthPlace	Place of birth	"Warsaw"
P17	country	Sovereign nation	"United States"
P159	headquarters	Global corporate headquarters city	"Cupertino"
P112	founder	Original founders or creators	["Steve Jobs", "Steve Wozniak"]
P452	industry	Primary economic sector	"consumer electronics"
P1454	legalForm	Corporate entity classification	"corporation"
P571	inception	Founding or establishment date	"1976-04-01"
P36	capital	Capital city (for nations/states)	"Washington, D.C."
P1082	population	Headcount population metric	"8336817"
P50	author	Author or literary writer	"William Shakespeare"
P57	director	Film or theatrical director	"Christopher Nolan"
P175	performer	Musical artist or band	"The Beatles"
P176	manufacturer	Manufacturing entity	"Boeing"

🔗 Cross-database authority identifiers (externalIds object)

Property Code	Output Key	Authority Registry	Use Case
P214	viaf	Virtual International Authority File	Library cataloging & global authority control
P213	isni	International Standard Name Identifier	Universal creator & organization reconciliation
P496	orcid	Open Researcher and Contributor ID	Academic research paper attribution
P345	imdb	Internet Movie Database	Film, television, and entertainment credits
P2002	twitter	Twitter / X Handle	Official social identity verification
P2013	facebook	Facebook Page ID	Corporate branding & social footprint
P2003	instagram	Instagram Handle	Influencer and corporate verification
P2397	youtube	YouTube Channel ID	Official video broadcast attribution
P4264	linkedinCompany	LinkedIn Corporate Handle	B2B lead generation & business intelligence
P6634	linkedin	LinkedIn Personal Profile	Executive and founder profiling
P227	gnd	Integrated Authority File (Germany)	European cultural & historical archiving
P244	loc	Library of Congress Control Number	US national library cataloging
P434	musicbrainz	MusicBrainz Artist ID	Music industry rights and metadata
P1902	spotify	Spotify Artist URI	Streaming catalog integration
P2427	grid	Global Research Identifier Database	Institutional research benchmarking

---

🏗️ Architecture & resolution flow

sequenceDiagram
    autonumber
    actor User as User / AI Agent
    participant Actor as Wikidata Entity Resolver
    participant Search as wbsearchentities
    participant Entities as wbgetentities
    participant Labels as wbgetentities (Fact Labels)

    User->>Actor: Seed names or Q-IDs
    Actor->>Search: Bounded concurrent lookup (8 parallel workers)
    Search-->>Actor: Matched Q-IDs (e.g. Q312, Q7186)
    Actor->>Entities: Batched fetch (chunks of 50 Q-IDs)
    Entities-->>Actor: Raw claims, labels, sitelinks, external IDs
    Actor->>Labels: Resolve referenced fact Q-IDs in one batch
    Labels-->>Actor: Readable strings (e.g. "physicist", "Cupertino")
    Actor->>User: Structured, normalized JSON dataset

---

🎯 Use cases & industry applications

1. AI agent grounding & hallucination prevention

• Entity linking in prompts: Before answering complex user questions, have your AI agent resolve mentioned names to verify factual founding dates, parents, or headquarters.
• RAG chunk enrichment: Prepend canonical Wikidata metadata (e.g., entity type, industry, aliases) to vector search embeddings to drastically improve retrieval accuracy.

2. Master Data Management (MDM) & entity reconciliation

• Customer & vendor deduplication: Standardize company names in your CRM (e.g. "Google LLC", "Google Inc", "Alphabet") to a single canonical entity (Q95).
• Cross-system identifier mapping: Automatically enrich internal customer profiles with LEI, VIAF, ISNI, and Wikidata identifiers.

3. Academic research & digital humanities

• Author & researcher profiling: Reconcile scholar names against ORCID, VIAF, and Library of Congress identifiers.
• Biographical knowledge graphs: Extract birthplaces, citizenships, and historical eras for large cohorts of historical figures.

4. Due diligence, KYC & OSINT

• Corporate footprint verification: Verify official corporate websites, executive leadership, legal form, and registered headquarters.
• Social media attribution: Collect verified brand social handles (Twitter, LinkedIn, YouTube) linked directly in Wikidata.

---

📥 Input parameters

The actor is designed to execute immediately with zero required configuration, falling back to a curated multi-domain seed dataset of 100+ entities when run empty.

Parameter	Type	Default	Description
queries	array	["Elon Musk", ...]	List of entity names, people, brands, places, or concepts to resolve (one per line).
language	select	"en"	Language code for labels, descriptions, and facts (en, de, fr, es, it, tr, ja, zh, etc.).
includeFacts	boolean	true	When true, pulls and resolves readable key facts (type, occupation, dates, country, founders, etc.).
maxResults	integer	0	Maximum entities to return (0 = unlimited, processes entire input list).
qids	array	[]	Optional list of specific Wikidata Q-IDs (e.g. Q317521, Q95) to bypass search and fetch directly.
proxyConfiguration	object	OFF	Proxy configuration. Disabled by default because Wikidata is open and keyless.

Sample JSON configuration

{
  "queries": [
    "OpenAI",
    "Jensen Huang",
    "Kyoto",
    "CRISPR gene editing",
    "Anthropic"
  ],
  "language": "en",
  "includeFacts": true,
  "maxResults": 100
}

---

📊 Output data fields

Every record pushed to the Apify dataset represents an authoritative resolved entity. Output fields use clean, normalized strings for effortless database ingestion.

Field	Coverage	Type	Description
query	🟢 High (100%)	String	Original search query or Q-ID submitted in input.
qid	🟢 High (100%)	String	Canonical permanent Wikidata Q-ID (e.g., "Q317521").
label	🟢 High (98%)	String	Official primary name in the requested language.
description	🟢 High (98%)	String	Concise explanatory summary statement.
type	🟢 High (96%)	String	Humanized entity classification (e.g., "human", "business").
aliases	🟡 Conditional (70%)	String	Comma-separated alternative names and acronyms.
officialWebsite	🟡 Conditional (60%)	String	Verified primary external website URL.
wikipediaUrl	🟢 High (92%)	String	Direct canonical link to the corresponding Wikipedia article.
wikidataUrl	🟢 High (100%)	String	Canonical link to the Wikidata knowledge graph entry.
facts	🟢 High (95%)	String	Serialized JSON object containing resolved fact properties.
externalIds	🟢 High (90%)	String	Serialized JSON object of cross-database authority IDs.
scrapedAt	🟢 High (100%)	String	ISO 8601 UTC execution timestamp.

---

📋 Example JSON outputs

1. Corporate technology entity

{
  "query": "Apple Inc",
  "qid": "Q312",
  "label": "Apple",
  "description": "American multinational technology company",
  "type": "technology company",
  "aliases": "Apple Inc., Apple Computer, Apple Computer, Inc., AAPL",
  "officialWebsite": "https://www.apple.com/",
  "wikipediaUrl": "https://en.wikipedia.org/wiki/Apple_Inc.",
  "wikidataUrl": "https://www.wikidata.org/wiki/Q312",
  "facts": "{\"instanceOf\":\"technology company\",\"country\":\"United States\",\"headquarters\":\"Cupertino\",\"founder\":[\"Steve Jobs\",\"Steve Wozniak\",\"Ronald Wayne\"],\"industry\":\"consumer electronics\",\"legalForm\":\"corporation\",\"inception\":\"1976-04-01\"}",
  "externalIds": "{\"viaf\":\"154701235\",\"isni\":\"0000000122982463\",\"twitter\":\"Apple\",\"youtube\":\"Apple\",\"linkedinCompany\":\"apple\",\"freebase\":\"/m/0k8z\"}",
  "scrapedAt": "2026-11-10T14:35:10.124Z"
}

2. Historical & scientific figure

{
  "query": "Marie Curie",
  "qid": "Q7186",
  "label": "Marie Curie",
  "description": "Polish-French physicist and chemist (1867–1934)",
  "type": "human",
  "aliases": "Maria Skłodowska-Curie, Madame Curie",
  "officialWebsite": null,
  "wikipediaUrl": "https://en.wikipedia.org/wiki/Marie_Curie",
  "wikidataUrl": "https://www.wikidata.org/wiki/Q7186",
  "facts": "{\"instanceOf\":\"human\",\"occupation\":[\"physicist\",\"chemist\",\"professor\"],\"birthDate\":\"1867-11-07\",\"deathDate\":\"1934-07-04\",\"citizenship\":[\"Poland\",\"France\"],\"birthPlace\":\"Warsaw\"}",
  "externalIds": "{\"viaf\":\"102353774\",\"isni\":\"0000000121459207\",\"gnd\":\"118523023\",\"imdb\":\"nm1517454\",\"loc\":\"n79042211\"}",
  "scrapedAt": "2026-11-10T14:35:11.890Z"
}

---

🤖 AI Agent & MCP integration guide

Every Apify Actor can be invoked as a tool by LLM agents via the Model Context Protocol (MCP).

Claude Desktop MCP configuration (claude_desktop_config.json)

{
  "mcpServers": {
    "apify": {
      "command": "npx",
      "args": [
        "-y",
        "@apify/mcp-server",
        "--actors",
        "haketa/wikidata-resolver"
      ],
      "env": {
        "APIFY_TOKEN": "YOUR_APIFY_TOKEN"
      }
    }
  }
}

Python LangChain tool example

from langchain_community.tools.apify import ApifyActorsTool
from apify_client import ApifyClient

client = ApifyClient("YOUR_APIFY_TOKEN")
run_input = {
    "queries": ["DeepSeek", "Mistral AI", "Ilya Sutskever"],
    "language": "en",
    "includeFacts": True
}

run = client.actor("haketa/wikidata-resolver").call(run_input=run_input)
for item in client.dataset(run["defaultDatasetId"]).iterate_items():
    print(f"[{item['qid']}] {item['label']}: {item['description']}")

Node.js integration

import { ApifyClient } from 'apify-client';

const client = new ApifyClient({
    token: 'YOUR_APIFY_TOKEN',
});

const run = await client.actor('haketa/wikidata-resolver').call({
    queries: ['Quantum computing', 'CRISPR', 'James Webb Space Telescope'],
    includeFacts: true,
});

const { items } = await client.dataset(run.defaultDatasetId).listItems();
console.log(`Resolved ${items.length} entities:`, items.map(i => i.label));

---

🚀 How to use this actor

1. Enter entity names: Enter one or more names, brands, or places into the Names to resolve input field.
2. Choose language: Select your desired target language (en is recommended for maximum claim coverage).
3. Configure fact extraction: Keep Include key facts enabled to translate internal Q-ID codes into readable strings.
4. Execute: Click Start. The actor processes requests in bounded concurrent batches, logging live entity resolution progress.
5. Export results: Download your clean, structured entity records in JSON, CSV, or Excel formats.

---

💡 Best practices & optimization tips

• Exact Q-ID targeting: When resolving unambiguous entities, paste the exact Q-ID (e.g. Q95) into the Q-IDs input to skip fuzzy search and fetch canonical records instantly.
• Handling JSON string fields: Fields such as facts and externalIds are stored as JSON strings to maintain tabular CSV/Excel compatibility. Parse them using JSON.parse() in JS or json.loads() in Python.
• Language fallback mechanism: When an entity statement lacks a translation in your requested language, the actor automatically falls back to English to prevent empty data fields.
• Zero proxy overhead: Do not enable proxies unless operating in a restricted enterprise network. Wikidata endpoints operate cleanly over direct HTTP.

---

❓ Frequently Asked Questions (FAQ)

Do I need a Wikidata or Wikimedia account?

No. The resolver accesses the public Wikidata MediaWiki API without requiring authentication, login tokens, or subscription plans.

How many names can I resolve in a single run?

There is no hard limit. The actor utilizes bounded concurrency pools that throttle searches and batch entity requests up to 50 at a time, easily processing batches of 1,000+ entities without triggering rate limits.

Are fact properties returned as codes or readable words?

All supported facts (occupation, country, headquarters, founder, industry, etc.) are automatically resolved to human-readable labels rather than raw property codes.

What happens if a searched name has no match?

If Wikidata returns no matching entity for a query, the actor logs a friendly warning and continues processing the remaining names in your queue without failing.

---

🔗 Related AI agent & data enrichment tools

• SEC EDGAR Scraper — Extract US public company filings and financials.
• Google Trends Scraper — Search interest over time and breakout query intelligence.
• G2 Reviews Scraper — Extract B2B software reviews and competitor pain points.
• Article Content Extractor — Clean text and metadata extraction from any article URL.

---

📋 Changelog

Version 1.2 (2026-11)

• Unlocked 1,000+ batch entity resolution: Expanded worker queues to seamlessly process 1,000+ names with bounded request fan-out.
• Expanded 100+ entity fallback knowledge base: Enriched zero-input demo mode with 100 diverse global entities spanning tech, history, geography, and science.
• Accelerated prefill execution: Streamlined batch entity fetching to resolve standard batches in seconds.
• Enhanced fact label resolution: Improved multi-lingual label fallback mechanics for foreign-language claims.
• Schema & MCP optimization: Verified full Apify CLI 2026 schema standards for zero-friction MCP agent tool calling.

Version 1.1 (2026-07)

• Added cross-database identifier extraction (VIAF, ISNI, ORCID, IMDb, GND).
• Implemented HTTP Standby mode for instant low-latency REST queries.
• Added support for direct Q-ID array input.

Version 1.0 (2026-07)

• Initial release featuring wbsearchentities search, wbgetentities batching, and claim parsing.