Google News MCP Server for AI Agents

An MCP (Model Context Protocol) server that gives your AI agent one tool: live news search over Google News. Ask a question, get current headlines back in the same turn — no API key, no scraping setup, no browser, no monthly plan.

This Actor runs in Apify Standby mode: it is a web server at a fixed URL that speaks MCP over streamable HTTP. Point Claude Desktop, Cursor, VS Code, n8n, LangChain or your own agent at it and search_news appears in its tool list.

You bring the model; we serve the data. This server performs no inference of its own, so you are never paying for somebody else's tokens on top of your own.

The tool

search_news

Argument	Type	Description
query	string	Keyword or phrase, e.g. "openai funding". Google News search operators work: when:1d, site:reuters.com, -word, OR, "exact phrase".
topic	enum	Browse a section instead of searching: WORLD, NATION, BUSINESS, TECHNOLOGY, ENTERTAINMENT, SCIENCE, SPORTS, HEALTH, POLITICS, ECONOMY, REAL_ESTATE, JOBS, EDUCATION, AUTOS, MOVIES, MUSIC, CELEBRITIES, ARTS, SOCCER, BASKETBALL.
limit	number	Max articles (default 10, max 100).
language	string	e.g. "en-US", "de", "fr" (default "en-US").
country	string	Edition, e.g. "US", "GB", "DE" (default "US").

Pass either query or topic, not both — Google serves topic sections from a different feed that ignores keywords, so accepting both would silently drop your query and hand your agent a plausible-looking wrong answer. The tool refuses instead, and does not charge.

Sample output

A real response to search_news({ query: "openai", limit: 2 }):

[
  {
    "title": "OpenAI kicks off charm offensive to change voters' view of data centers",
    "publisher": "Reuters",
    "publisherDomain": "reuters.com",
    "url": "https://news.google.com/rss/articles/CBMirgFBVV95cUxNT3pvX1Vf...",
    "publishedAt": "2026-11-05T19:34:01.000Z",
    "relatedCoverage": [
      { "title": "OpenAI pledges to halt NDAs in new data center policy proposal after election rebuke", "publisher": "CNBC" },
      { "title": "OpenAI Pledges to Give Locals More Say in Data Center Projects", "publisher": "The New York Times" }
    ]
  }
]

relatedCoverage is the field most agents end up leaning on: it is the same story as reported by other publishers, which is what you need to tell a real consensus from one outlet's framing.

Pricing and how we compare

$0.005 per search_news call that returns at least one article — flat, however many articles it returns, and no Actor-start fee.

No result, no charge. A search with zero matches, a call rejected for a bad or missing argument, and a call where Google News itself did not answer are all free. You are billed for answers, not attempts.

A 100-article call costs the same $0.005 as a 1-article call, so set limit to what you actually want to read.

Prices read live from each Actor's own pricingInfos record (checked 2026-11-05, re-confirmed on the latest news mcp / news mcp server / google news mcp queries):

Actor	Users (u30d)	Closest event	Price/call	vs. ours
fetchsmith/google-news-mcp-server	—	search_news	$0.005	—
mrbridge/latest-news-mcp-server	118 (35)	tool-read	$0.008 (FREE) tiered down to $0.005 (GOLD+), plus a second tool-enriched event at $0.015→$0.008 and a $0.00005 start fee	matches our price only at its best tier; 60% dearer on a FREE-plan caller, plus a start fee we don't charge
nexgendata/news-mcp-server	20 (4)	tool-call	$0.02 + $0.00005 start fee	4x dearer, plus a start fee we don't charge
quodlibetical_buffalo/google-news-mcp	2 (1)	— (no PPE pricing record found)	—	direct title collision, but no measurable traction or disclosed price to compare

mrbridge/latest-news-mcp-server is the clear demand leader in this corner (u30d=35, second only to ESPN across every MCP listing we've measured) and the one worth beating: our flat $0.005 beats its entry-tier FREE-plan price by 37% and never requires a start fee or a second enrichment charge to reach parity. nexgendata/news-mcp-server is a flat 4x dearer. Neither offers relatedCoverage (what other publishers ran the same story) or a topic browse mode — both are query-only.

Use cases

• News-aware agents — let an assistant check what actually happened today before it answers, instead of asserting from stale training data.
• Monitoring and alerting — poll search_news({ query: "<your brand> when:1d" }) on a schedule and have the agent decide whether anything is worth escalating.
• Research and briefings — pull a topic section, then use relatedCoverage to see which stories several outlets independently picked up.
• Market and competitor tracking — site: and - operators narrow a query to the publications you trust.
• Sentiment and trend inputs — headlines plus publishers plus timestamps, structured, for a downstream model.

How to connect

Standby URL:

https://fetchsmith--google-news-mcp-server.apify.actor/mcp

Add it as a streamable-HTTP MCP server, with your Apify token as a bearer credential. In a mcpServers config this looks like:

{
  "mcpServers": {
    "google-news": {
      "url": "https://fetchsmith--google-news-mcp-server.apify.actor/mcp?token=YOUR_APIFY_TOKEN"
    }
  }
}

A raw check, if you want to see it answer before wiring an agent up:

curl -s -X POST "https://fetchsmith--google-news-mcp-server.apify.actor/mcp?token=YOUR_APIFY_TOKEN" \
  -H 'Content-Type: application/json' \
  -H 'Accept: application/json, text/event-stream' \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"search_news","arguments":{"query":"openai","limit":2}}}'

FAQ

Is this legal / is it scraping? No scraping and nothing behind a login. It reads Google News' own public RSS feeds — the same endpoints Google publishes for feed readers.

Why is url a news.google.com link instead of the publisher's own URL? Because resolving Google's redirect costs an extra HTTP request per article and Google rate-limits it. An MCP tool call is synchronous — your agent is waiting — so this server keeps the round trip short. The link works: it opens the article. If you need resolved publisher URLs and full article text in bulk, that is what the batch Actor below is for.

Do I need an OpenAI/Anthropic key? No. This server returns data. Your agent already has whatever model it uses.

How fresh is the data? Live. Every call hits Google News at that moment; nothing is cached on our side.

Can I search non-English news? Yes — set language and country (e.g. language: "de", country: "DE").

What if I need thousands of articles, full text, or a CSV? Use the batch Actor: google-news-scraper — pay per article, resolved URLs, full article bodies, date windows, ticker extraction.

Related guides

• fetchsmith.com/tools
• google-news-scraper — the batch version: full article text, resolved URLs, CSV export
• Standby MCP — one URL that's always on, instead of an Actor run you poll
• An MCP tool that refuses bad calls instead of guessing — and doesn't charge for the refusal
• Google News RSS: public feeds, no key
• Why the Google News RSS description is not a summary
• Decoding Google News RSS redirect links