Newsmcp

What pranciskus/newsmcp MCP server does

The pranciskus/newsmcp MCP server gives AI agents access to current world-news events rather than isolated article searches. Its data is organized around events: related articles are grouped together, then labeled by topic and geographic region. Events include fields such as summaries, topic and geography tags, source counts, article counts, timestamps, and impact scores.

Coverage is organized into 12 topics, including politics, economy, technology, science, health, environment, sports, crime, military, education, and society. Regional filters cover six continents and 24 countries. The available region list should be retrieved through the server when an agent needs the exact supported slugs.

How it works

Hundreds of news sources are continuously collected and processed. Articles about the same event are grouped using vector embeddings. The resulting events are classified with topic and geographic entity tags, then ranked using source count, impact, and recency. As new articles arrive, events can be updated and duplicate reporting is merged.

The MCP process serves the same underlying data as the REST API at https://newsmcp.io/v1. The REST endpoints provide event lists, individual event details, topic categories, and region categories, while the MCP interface makes these operations available as agent tools.

Setup and configuration

Install the server with npm and configure the MCP client to run npx -y @newsmcp/server. The README provides configuration examples for Claude Desktop, Cursor, Windsurf, Claude Code, OpenAI Codex, Gemini CLI, and OpenCode. No API key or additional credential is required for the default backend.

The default API backend can be replaced by setting NEWS_API_BASE_URL in the MCP server environment. The documented example points this variable to an alternative backend URL ending in /v1; the variable is optional when using the hosted default service.

Tools and capabilities

The pranciskus/newsmcp MCP server exposes four tools:

• get_news returns current events with optional topic, region, and time-window filters. It supports pagination and sorting by source count, impact score, last-seen time, or article count. The time window accepts 1 to 168 hours, and results are limited to a maximum of 50 per page.
• get_news_detail accepts an event UUID from get_news and returns the event's source articles, generated context, impact analysis, and entity tags.
• get_topics returns all available topic categories without parameters.
• get_regions returns the supported geographic regions without parameters.

Limitations and notes

The service focuses on aggregated news events and their associated source articles. It does not require an API key, but its usefulness depends on the current data supplied by the hosted backend. Filters must use the supported topic and region slugs, and event detail lookup requires an event ID obtained from a news result. The repository also documents direct REST access for clients that do not use MCP.