News data for AI agents: a news MCP server

AI agents are great at reasoning and terrible at knowing what just happened. The Model Context Protocol fixes that — and we've shipped an MCP server that gives any agent live, already-scored news in five tools.

The problem MCP solves

An agent that can't see current events hallucinates them. Bolting on a raw news API helps, but then the agent has to score, classify and de-duplicate everything itself — inside a reasoning loop, badly. The Model Context Protocol lets you hand the agent tools instead of raw endpoints.

What the server gives your agent

Our MCP server wraps the NewsAgent Data API and exposes five tools — every article they return already carries an urgency score (0-10), political-lean label, topic, country and audience tags:

• get_feed — filter by country, topic, language, lean, audience, urgency, date
• search_news — full-text keyword search
• get_breaking — recent high-urgency items
• coverage_stats — live totals (no key)
• list_sources — source catalog (Standard tier)

Setup is two minutes

Grab a free key, download the server, and add four lines to your client config:

{
  "mcpServers": {
    "newsagent": {
      "command": "python",
      "args": ["/path/to/newsagent_mcp.py"],
      "env": { "NEWSAGENT_API_KEY": "your_key" }
    }
  }
}

Restart, and the tools appear. Now you can just ask: "Get breaking sanctions news from Russia scored 7+," or "Compare how state vs opposition sources framed Ukraine this week." The agent picks the tool, applies the filters, and gets back classified results.

Why it matters

No other news API ships an MCP server. If you're building agents, monitoring pipelines or RAG systems, this is the shortest path from "the agent needs current events" to "the agent has scored, filtered news" — with no NLP layer to build. Full setup on the MCP page.