Home / News API for AI agents

News API for AI agents: current news without long articles

A news API for AI agents that returns short fact records instead of pages of text. Connect through the MCP server or plain REST, and keep your context window for what matters.

Why compact

Facts fit in a context window

Each story is a title, a few fact lines, numbers with units, people, organizations and places. Coverage from many outlets is merged, so your agent reads each story once.

Short records

Fact lines instead of paragraphs. Use facts=top3 and fields to cut each record further.

Stable IDs

Every story keeps its ID. Embed it, store the ID as metadata, and refresh with updated_since.

Numbers as data

Numbers carry a value and a unit code such as percent or USD. Search them with /v1/facts.

MCP server

Connect your agent in one config

The server runs at https://api.latestnewsapi.com/mcp over streamable HTTP and uses the same X-API-Key header as REST.

ToolUse it to
search_storiesSearch and filter stories by keyword, category, person, organization, place, date and importance.
get_storyFetch one story with all its fact lines.
trendingList the most-covered stories in a recent window.
search_factsFind numbers by subject, unit code and value range.
countCount stories over time or by facet.

One tool call is 1 credit. Connecting and listing tools are free.

mcp client config (example)
{
  "mcpServers": {
    "latestnewsapi": {
      "url": "https://api.latestnewsapi.com/mcp",
      "headers": { "X-API-Key": "YOUR_KEY" }
    }
  }
}

Config file names differ by client. The URL and header are the same everywhere.

RAG

Compact facts for retrieval

Embed the title plus the fact lines, keep the ID as metadata, and use relative dates like from=24h with min_importance=60 for a "what matters now" query.

request
curl -H "X-API-Key: YOUR_KEY" \
  "https://api.latestnewsapi.com/v1/stories?q=inflation&fields=id,title,facts,category,importance,outlet_count&facts=top3&limit=1"
response 200
{
  "id": 13023,
  "title": "Global economic outlook projects lower oil prices and continued monetary tightening",
  "facts": [
    "S&P Global expects Brent crude oil prices to drop to 70 dollars by the end of 2028.",
    "The Global Economic Outlook Report for November 2026 was prepared by Ken Wattret, Chief Economist at S&P Global Market Intelligence.",
    "The report stated energy prices are expected to decline as Middle East conflicts have a reduced impact on oil supply."
  ],
  "category": "finance",
  "importance": 48,
  "outlet_count": 1
}
Tool definitions

OpenAPI as your tool list

Load the spec and turn each operation into a function-calling tool for your framework. Parameters, types and allowed values are all in it.

python
import requests

URL = "https://api.latestnewsapi.com/openapi.json"
spec = requests.get(URL).json()
for path, item in spec["paths"].items():
    for method, op in item.items():
        print(method.upper(), path)

/openapi.json

The full spec: every endpoint, filter and error.

/llms.txt

A plain-text map of the API for agents and tooling.

/docs.md

The whole documentation as one Markdown file.

Reliability

Errors an agent can fix

Every error has one shape. For invalid parameters it lists the allowed values, so the agent can read them and retry without a human.

422 invalid_parameter (shape)
{
  "error": {
    "code": "invalid_parameter",
    "status": 422,
    "param": "sort",
    "invalid_values": ["newest"],
    "allowed_values": ["-first_published", "-updated", "-importance", "-coverage"],
    "request_id": "req_..."
  }
}
Facts and our own titles, never copied article text. Your agent can show an answer built from these facts without reproducing anyone else's writing.

Plan limits return 403 plan_restriction in the same format. Failed calls cost 0 credits.

FAQ

Questions

Does LatestNewsAPI have an MCP server?

Yes, at https://api.latestnewsapi.com/mcp over streamable HTTP. It has five tools: search_stories, get_story, trending, search_facts and count. It uses the same API key header as the REST API.

What does an MCP call cost?

One tool call costs 1 credit. Connecting and listing the tools cost 0 credits. Failed calls cost 0.

Why compact facts instead of full articles for RAG?

Each story is a short record with a title, fact lines and a stable ID. More stories fit in a context window, and you can refresh one record when it changes.

Can my agent show the facts to users?

Titles and fact lines are in our own wording. We never return copied article text, so they are meant to be displayed. See the terms for plan rules.

Can I use the OpenAPI file as a tool definition?

Yes. The spec at /openapi.json lists every endpoint and parameter, so you can turn it into function-calling tools for your framework.

Is there an llms.txt file?

Yes, at /llms.txt. It describes the API in plain text, including endpoints, filters and error shapes. The docs are also available as Markdown at /docs.md.

Give your agent the news

Start on the free plan: 100 requests a day, no credit card.