Agentic Google News API

agentic-google-news-api is a compact Google News search API for AI agents, MCP workflows, automations, brand monitoring, and market monitoring. It reads Google's public RSS search feeds with ordinary HTTP requests—no browser automation, proxies, LLMs, or paid data providers.

Each article becomes one normalized default-dataset item, so an agent can retrieve JSON, JSONL, CSV, or the Apify dataset API directly after a run.

What it does

• Searches one to twenty Google News queries in a selected country and language.
• Supports optional from and to dates in YYYY-MM-DD.
• Keeps at most 100 valid records per query.
• Removes duplicate articles deterministically by retaining the first occurrence in input-query and feed order.
• Isolates a failed query so successful queries still return results.
• Uses bounded concurrency, request timeouts, and exponential retries only for transient failures.

It does not scrape full articles or use browser automation. articleUrl is null unless the RSS record already provides a direct non-Google-News URL; googleNewsUrl is always the RSS link.

Input

{
  "queries": ["OpenAI", "AI regulation"],
  "country": "US",
  "language": "en",
  "maxResultsPerQuery": 50,
  "from": "2026-08-01",
  "to": "2026-08-14",
  "deduplicate": true
}

queries is required and accepts 1–20 non-empty strings. country defaults to US, language to en, maxResultsPerQuery to 100 (and cannot exceed it), and deduplicate to true.

Output

Every dataset item has this stable machine-readable shape:

{
  "query": "OpenAI",
  "title": "Example headline",
  "source": "Example Publisher",
  "sourceDomain": "example.com",
  "publishedAt": "2026-08-14T08:30:00.000Z",
  "googleNewsUrl": "https://news.google.com/rss/articles/...",
  "articleUrl": null,
  "language": "en",
  "country": "US",
  "fetchedAt": "2026-08-14T09:00:00.000Z"
}

Nullable fields are explicitly null when Google omits them or their value is not reliable. All usable dates are normalized to ISO-8601 UTC.

Pay-per-event and agentic payments

The extraction path calls Actor.pushData(article, 'news-result') once per retained article. On Apify, this is the official atomic dataset-item/pay-per-event path: records that are discarded as duplicates, malformed, from a failed query, or absent are never charged. If the caller's event limit is reached, the Actor stops accepting additional results.

Set the price for the news-result event in Apify Console after deployment. Pricing stays out of extraction code. An Actor using pay per event, limited permissions, and no Standby mode meets the current technical conditions for Apify agentic payments such as x402; final Store eligibility is determined by Apify after configuration and publication. See DEPLOY.md.

Local development

Requires Node.js 20 or later.

npm install
npm test
npm run build

To run locally, place an INPUT.json file in storage/key_value_stores/default/ (Apify local-storage convention), then run:

$npm start

To validate the live public RSS endpoint without writing data:

$npm run smoke

To exercise local pay-per-event accounting, run the Actor with ACTOR_TEST_PAY_PER_EVENT=true; Apify writes its local charging events to storage/datasets/charging-log/.

Deployment

The repository includes the Actor manifest, schemas, Dockerfile, source, tests, and deploy guidance. Install the Apify CLI, authenticate it, and follow DEPLOY.md. The intended Store positioning is Google News API, Google News search API, real-time news data, news search for AI agents, brand monitoring, market monitoring, MCP workflows, and x402/agentic payments.

Operational notes

Google News RSS is a public external source and its coverage, links, and query behavior may change. The Actor deliberately returns only RSS metadata and never claims complete coverage or direct publisher URLs it has not reliably received.