Build Your Own News Hub: Ask Claude \"What Did Elon Musk Do Today?\"

Imagine asking Claude, "What did Elon Musk do in the last 24 hours?" and getting a comprehensive answer based on real-time news from Google News, Reddit, and Hacker News. Today we'll show you how to build this in under 10 minutes with OpenAgents.

The Vision

We've created a new demo that showcases the power of combining:

• The Feed Mod - Immutable news broadcasting with full-text search
• RSS/API Collection - Automatic news gathering from multiple sources
• MCP Integration - Connect any LLM (like Claude Desktop) to query your feed

The result? Your own personal news hub that any AI assistant can tap into.

What You'll Build

┌─────────────────────────────────────────────────────────────┐
│                    OpenAgents Network                        │
│                                                              │
│  ┌──────────────┐     ┌──────────────┐                      │
│  │    News      │────>│   Feed Mod   │<────── MCP ─────────┐│
│  │  Collector   │     │   (Storage)  │                     ││
│  │    Agent     │     └──────────────┘                     ││
│  └──────────────┘            │                             ││
│         │                    │                             ││
│         ▼                    ▼                             ││
│  ┌──────────────┐     ┌──────────────┐                     ││
│  │ Google News  │     │  Searchable  │                     ││
│  │ Reddit RSS   │     │    Posts     │                     ││
│  │ Hacker News  │     │  + Tags      │                     ││
│  └──────────────┘     └──────────────┘                     ││
└─────────────────────────────────────────────────────────────┘│
                                                               │
┌──────────────────────────────────────────────────────────────┘
│
│   ┌──────────────────┐
└──>│  Claude Desktop  │ "What did Elon do today?"
    │    (via MCP)     │
    └──────────────────┘

Quick Start: 5-Minute Setup

Step 1: Install OpenAgents

# Clone the repository
git clone https://github.com/acenta-ai/openagents.git
cd openagents
 
# Install with pip (Python 3.10+ required)
pip install -e .

That's it! No Docker, no complex setup, just a simple pip install.

Step 2: Start the News Tracker

cd demos/06_elon_musk_tracker
 
# Option A: One command (runs both network and agent)
./run.sh
 
# Option B: Manual (two terminals)
# Terminal 1:
openagents network start network.yaml
 
# Terminal 2:
python agents/news_collector.py

You should see:

[NewsCollector] Connected! Starting news collection loop (interval: 300s)
[NewsCollector] Fetching news at 2025-12-18 10:30:00...
[NewsCollector] Found 15 new items to post.
[NewsCollector] Posted: Elon Musk's net worth soars, now more than double...
[NewsCollector] Posted: SpaceX Starship completes fourth test flight...
...

Step 3: Connect Claude Desktop

Add this to your Claude Desktop configuration:

macOS: ~/Library/Application Support/Claude/claude_desktop_config.json

Linux: ~/.config/claude/claude_desktop_config.json

Windows: %APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "elon-news": {
      "command": "npx",
      "args": ["mcp-remote", "http://localhost:8700/mcp"]
    }
  }
}

Or if you prefer using curl for a quick test:

# List available tools
curl -X POST http://localhost:8700/mcp \
  -H "Content-Type: application/json" \
  -d '{"method": "tools/list", "params": {}}'

Step 4: Ask Claude!

Open Claude Desktop and try:

"What did Elon Musk do in the last 24 hours?"

Claude will use the feed search tools to query your local news hub and provide a comprehensive summary.

How It Works

The News Collector Agent

The NewsCollectorAgent runs continuously, fetching news every 5 minutes from:

Each news item is automatically tagged:

# Auto-categorization based on content
tags = ["elon-musk", "news"]
if "tesla" in content: tags.append("tesla")
if "spacex" in content: tags.append("spacex")
if "twitter" in content or "x.com" in content: tags.append("x-twitter")
# etc...

The Feed Mod

Posts are stored immutably with full-text search:

# Agent creates a post
await feed_adapter.create_post(
    title="SpaceX Starship Launch Success",
    content="Full details about the launch...",
    tags=["spacex", "starship", "launch"]
)
 
# Claude queries the feed via MCP
await feed_adapter.search_posts(query="Starship launch", limit=10)
await feed_adapter.get_recent_posts(since_timestamp=yesterday)

MCP Integration

The network exposes these tools via MCP:

When Claude asks about Elon Musk, it can call these tools to search your collected news.

Customizing Your News Hub

Track Different Topics

Edit tools/rss_fetcher.py to add new sources:

def fetch_crypto_news():
    """Fetch cryptocurrency news."""
    return fetch_google_news_rss("Bitcoin OR Ethereum", count=10)
 
def fetch_ai_news():
    """Fetch AI news."""
    return fetch_google_news_rss("OpenAI OR Anthropic OR Google AI", count=10)

Change Collection Frequency

# Fetch every minute for real-time news
python agents/news_collector.py --interval 60
 
# Fetch every hour for less frequent updates
python agents/news_collector.py --interval 3600

Add Authentication

Secure your MCP endpoint by adding to network.yaml:

external_access:
  auth_token: "your-secret-token"
  # Or use environment variable
  auth_token_env: "NEWS_HUB_TOKEN"

Then include the token in your MCP client:

{
  "mcpServers": {
    "elon-news": {
      "command": "npx",
      "args": ["mcp-remote", "http://localhost:8700/mcp"],
      "env": {
        "AUTHORIZATION": "Bearer your-secret-token"
      }
    }
  }
}

Deploy to Production

Run 24/7 with systemd

Create /etc/systemd/system/elon-news-tracker.service:

[Unit]
Description=Elon Musk News Tracker
After=network.target
 
[Service]
Type=simple
User=youruser
WorkingDirectory=/path/to/openagents/demos/06_elon_musk_tracker
ExecStart=/bin/bash run.sh
Restart=always
 
[Install]
WantedBy=multi-user.target

sudo systemctl enable elon-news-tracker
sudo systemctl start elon-news-tracker

Expose to the Internet

Use a reverse proxy like nginx:

server {
    listen 443 ssl;
    server_name news.yourdomain.com;
 
    location /mcp {
        proxy_pass http://localhost:8700/mcp;
        proxy_http_version 1.1;
        proxy_set_header Upgrade $http_upgrade;
        proxy_set_header Connection "upgrade";
    }
}

Now anyone with your MCP endpoint can connect their Claude to your news hub!

Real-World Example Queries

Once set up, try these prompts with Claude:

• "What did Elon Musk do in the last 24 hours?"
• "Summarize the latest SpaceX news"
• "What are people on Reddit saying about Tesla?"
• "Find any news about Neuralink from today"
• "What's trending about Elon on Hacker News?"

Why Build This?

1. Real-time knowledge: Your AI assistant gets fresh news, not stale training data
2. Privacy: News stays on your machine, not sent to external services
3. Customizable: Track any topic, from any source
4. Shareable: Expose via MCP to any AI assistant that supports it

What's Next?

We're working on:

• More news sources: RSS aggregator configuration
• Sentiment analysis: Auto-tag positive/negative news
• Topic clustering: Group related stories automatically
• Notifications: Alert when specific topics trend

Try It Today

The Elon Musk News Tracker demo is available now in demos/06_elon_musk_tracker/. Clone the repo, run the script, and start asking Claude about the latest news!

Have questions or feedback?

• Join our Discord community
• Open an issue on GitHub
• Follow us on Twitter

Happy tracking!

The OpenAgents Team

Changelog

Demo 06: Elon Musk News Tracker

New Files:

• demos/06_elon_musk_tracker/network.yaml - Network config with feed mod
• demos/06_elon_musk_tracker/agents/elon_fan_agent.py - News collection agent
• demos/06_elon_musk_tracker/tools/rss_fetcher.py - RSS/API fetching utilities
• demos/06_elon_musk_tracker/run.sh - One-command demo runner
• demos/06_elon_musk_tracker/README.md - Demo documentation

Features:

• Multi-source news aggregation (Google News, Reddit, Hacker News)
• Automatic topic categorization (tesla, spacex, x-twitter, etc.)
• Feed mod integration for immutable post storage
• MCP endpoint for external agent access
• Zero extra dependencies (uses only requests + stdlib)