Eloa - AI Content Curator

from Vaishnavi-2506

All your RSS feeds aggregated in one place. Eloa highlights what's new, shows the source, and how long ago it was published. Filter by read and unread.

🔥🔥FreeQuick setup

News Hub MCP

An MCP (Model Context Protocol) server that gives AI assistants real-time news fetching, search, NLP analysis, and personalized news preferences — all accessible through natural language.

Works with VS Code Copilot, Claude Desktop, Cursor, Windsurf, and any other MCP-compatible AI client.

What It Does

Ask your AI assistant things like:

"What's happening in the world today?" → fetches, scrapes, and summarizes top stories
"How are different outlets covering the Fed rate decision?" → side-by-side source comparison
"Search for Tesla news and analyze the sentiment" → tool chaining across search + NLP
"Set my preferences to technology news from India" → saved, applied to all future requests

Available Tools

Tool	Description
`get_news_digest`	Primary tool. Fetches, scrapes, and summarizes news in one call. Use for any general news question.
`get_latest_news`	Returns raw article metadata (titles, URLs, sources). Use when the user wants a list of links.
`search_news`	Searches articles by keyword or phrase. Use for specific names, events, or topics.
`compare_news_sources`	Fetches the same topic from 3 sources and returns a side-by-side summary of each.
`analyze_sentiment`	Scores text as positive, negative, or neutral using AFINN scoring.
`extract_entities`	Extracts people, organizations, and topics from text.
`get_my_preferences`	Returns your currently saved news preferences.
`set_my_preferences`	Updates your preferences (topics, country, language). Changes apply immediately.

Tool Chaining

The AI client can call tools in sequence. Example chains:

Sentiment analysis on news:

"Get me AI news and analyze the overall sentiment"

get_latest_news (topic: AI) → returns article descriptions
analyze_sentiment (text: combined descriptions) → returns score

Entity extraction from search:

"Search for Tesla news and extract key people and organizations"

search_news (query: Tesla) → returns articles
extract_entities (text: combined descriptions) → returns entities

User Preferences

Preferences are stored at ~/.news-hub-mcp/preferences.json and persist across sessions.

To set via wizard:

npm run setup

To set via AI:

"Set my news preferences to technology and science, country US, language English"

Preference fields:

Field	Description	Example
`topics`	Preferred topics, auto-applied when no topic given	`["technology", "science"]`
`country`	ISO 3166-1 alpha-2 country code	`"US"`, `"GB"`, `"IN"`
`language`	ISO 639-1 language code	`"en"`, `"fr"`, `"de"`

Development

npm start        # Run the server
npm run dev      # Run with hot reload (nodemon)
npm run setup    # Re-run the setup wizard

Architecture

AI Client (Copilot / Claude Desktop / Cursor)
        │
        │  MCP Protocol over stdio
        ▼
server/index.js          ← Entry point: validates env, registers tools, routes requests
        │
        ├── server/tools/          ← One file per MCP tool
        │   ├── getNewsDigest.js       Primary: fetch + scrape + summarize
        │   ├── getLatestNews.js       Raw article list
        │   ├── searchNews.js          Keyword search
        │   ├── compareNewsSources.js  Multi-source comparison
        │   ├── analyzeSentiment.js    NLP sentiment
        │   ├── extractEntities.js     NLP entity extraction
        │   ├── getMyPreferences.js    Read preferences
        │   └── setMyPreferences.js    Write preferences
        │
        └── server/services/       ← Reusable modules (no MCP coupling)
            ├── newsService.js         RapidAPI HTTP calls + response normalization
            ├── nlpService.js          sentiment + compromise NLP libraries
            ├── scraper.js             axios + cheerio article scraping
            └── preferencesService.js  ~/.news-hub-mcp/preferences.json read/write

Prerequisites

Node.js 18 or higher
A free RapidAPI account with the Real-Time News Data API subscribed
- Sign up at rapidapi.com
- Search for "Real-Time News Data" and subscribe to the free tier (150 req/month)
- Copy your x-rapidapi-key from the Endpoints tab

Installation

git clone https://github.com/your-username/news-hub-mcp.git
cd news-hub-mcp
npm install
npm run setup

npm run setup is an interactive wizard that:

Saves your RapidAPI key to .env and .vscode/mcp.json
Asks for your preferred topics, country, and language
Stores preferences at ~/.news-hub-mcp/preferences.json

MCP Client Configuration

VS Code Copilot

Add to .vscode/mcp.json in your workspace (created automatically by npm run setup):

{
  "servers": {
    "news-hub-mcp": {
      "type": "stdio",
      "command": "node",
      "args": ["server/index.js"],
      "cwd": "/absolute/path/to/news-hub-mcp",
      "env": {
        "RAPIDAPI_KEY": "your_key_here"
      }
    }
  }
}

Claude Desktop

Add to ~/Library/Application Support/Claude/claude_desktop_config.json:

{
  "mcpServers": {
    "news-hub-mcp": {
      "command": "node",
      "args": ["/absolute/path/to/news-hub-mcp/server/index.js"],
      "env": {
        "RAPIDAPI_KEY": "your_key_here"
      }
    }
  }
}