Wikipedia Search — Article Lookup to JSON

Search Wikipedia by keyword and get matching article titles, descriptions and snippets as JSON — $0.001 per result.

Agents grounding their answers need fast, structured access to encyclopedic facts. This actor turns a query into the matching Wikipedia articles with their short descriptions and snippets, so research and RAG agents can disambiguate entities and cite a stable source.

What this actor does

• Search Wikipedia by keyword for matching articles
• Each result: page title, short description, matching snippet (clean text), page key
• Open any article at en.wikipedia.org/wiki/{key}
• Batch many terms per run; cap spend with maxResults
• Agent-ready: flat JSON with sourceUrl + scrapedAt for citation

You only pay for successful results — failed or empty lookups cost nothing.

Why pick this Actor

• Clean entity grounding: title, stable page key, one-line description, and excerpt — disambiguation-ready for RAG pipelines
• Per-result pricing ($0.001/result) with a hard maxResults spend cap — empty lookups cost $0
• Flat, stable JSON schema with sourceUrl + scrapedAt on every item — citation-ready for RAG and grounding
• Batch many queries in one run; overlapping results are deduplicated and charged once
• MCP server, OpenAPI schema, and LangChain/CrewAI tool support out of the box — no glue code

Sample output

Each dataset item is flat, typed JSON with a sourceUrl and scrapedAt for citation/grounding:

{
  "query": "agentic commerce",
  "source": "wikipedia",
  "title": "Agentic commerce",
  "description": "Form of automated electronic commerce",
  "key": "Agentic_commerce",
  "sourceUrl": "https://en.wikipedia.org/w/rest.php/v1/search/page?q=agentic+commerce",
  "scrapedAt": "2026-06-11T15:00:00.000Z"
}

Input

{"queries":["agentic commerce"]}

Field	Type	Description
queries / query	array / string	Search term. One or many.
maxResults	integer	Hard spend cap (billed per result).
keywords / postedAfter	filters	Narrow results; enable delta/scheduled runs.

How much does it cost

Pay-per-result: $0.001 per successful result. No subscription, no compute-unit guesswork, no charge for empty results. An orchestrator can cap spend with maxResults.

How to use it with AI agents (MCP), Claude, and the API

Claude Desktop / Claude Code via Apify MCP

{
  "mcpServers": {
    "apify": {
      "command": "npx",
      "args": ["-y", "@apify/actors-mcp-server", "--actors", "oblanceolate_mandola/wikipedia-search"],
      "env": { "APIFY_TOKEN": "<YOUR_APIFY_TOKEN>" }
    }
  }
}

Python (Apify API)

from apify_client import ApifyClient
client = ApifyClient("<YOUR_APIFY_TOKEN>")
run = client.actor("oblanceolate_mandola/wikipedia-search").call(run_input={"queries":["agentic commerce"]})
for item in client.dataset(run["defaultDatasetId"]).iterate_items():
    print(item)

TypeScript (Apify API)

import { ApifyClient } from 'apify-client';
const client = new ApifyClient({ token: '<YOUR_APIFY_TOKEN>' });
const run = await client.actor('oblanceolate_mandola/wikipedia-search').call({"queries":["agentic commerce"]});
const { items } = await client.dataset(run.defaultDatasetId).listItems();
console.log(items);

LangChain / CrewAI tool

from langchain_apify import ApifyActorsTool
tool = ApifyActorsTool("oblanceolate_mandola/wikipedia-search")  # agent calls it autonomously

OpenAPI schema for self-integrating GPT agents is auto-exposed at the Actor's API tab.

Data & compliance

Reads only publicly accessible endpoints. No login, no credential harvesting, no CAPTCHA bypass. Every result carries its sourceUrl so downstream agents can cite and re-verify.

FAQ

How do I open the article?

Each result includes its page key; the article is at en.wikipedia.org/wiki/{key}.

Are snippets clean text?

Yes — search-match HTML in excerpts is stripped to readable text.

What is the source?

The official Wikimedia REST search API for English Wikipedia.

Can AI agents call this Actor directly?

Yes — via the Apify MCP server (snippet above), the OpenAPI schema on the Actor's API tab, or the LangChain/CrewAI tool wrapper. Results are flat JSON with sourceUrl and scrapedAt on every item, so downstream agents can cite and re-verify.

What happens when there are no results?

You pay nothing. Billing is per dataset item delivered, so an empty lookup costs $0, and the run log states why (no match, source rate limit) instead of failing silently.

Changelog

• 1.0 — Initial release: wikipedia.