MCP AI News Server

A Model Context Protocol (MCP) server that fetches and filters AI-related news from 20+ RSS feeds with scheduled notifications and email digests. Get your daily AI news delivered automatically via desktop notifications or email!

Features

• 20+ News Sources: TechCrunch, Hacker News, arXiv, Google AI, NVIDIA, VentureBeat, and more
• 📅 Scheduled Digests: Automatic daily/weekly news delivery at your preferred time
• 🔔 Desktop Notifications: Native system notifications with news summaries
• 📧 Email Digests: Beautiful HTML emails with latest AI news
• AI News Filtering: Automatically filters content for AI/ML-related keywords
• Category Selection: Choose from Tech, AI, Research, Community, or Hacker News categories
• Custom Source Selection: Pick specific sources for targeted news
• Keyword Search: Search across all sources for specific topics
• Parallel Fetching: Fast performance by fetching multiple feeds simultaneously
• MCP Integration: Works seamlessly with Claude Desktop
• Timezone Support: Configure for Australian Eastern or any timezone

Installation

# Install dependencies
npm install

# Build the project
npm run build

🚀 Quick Start: Notifications

Set Up Daily Digest

1. Enable desktop notifications
2. Configure schedule for 8 AM daily in Australia/Sydney
3. Send me a test digest now

That's it! You'll now get AI news every morning at 8 AM.

See NOTIFICATIONS.md for complete setup guide including email configuration.

News Sources

20 sources across 5 categories:

• Hacker News (2): Latest posts, Front page
• Tech News (6): TechCrunch, Ars Technica, The Verge, Wired, Engadget, ZDNet
• AI Publications (6): MIT Tech Review, VentureBeat, AI Trends, Google AI, NVIDIA, Microsoft AI
• Research (3): arXiv AI, arXiv ML, arXiv Computational Linguistics
• Community (3): Dev.to, Medium AI, Medium ML

See SOURCES.md for complete details on all sources and usage patterns.

Available Tools

1. get_ai_news

Fetches recent AI-related news from multiple RSS feeds with automatic filtering.

Parameters:

• limit (optional, default: 10): Maximum number of items to return
• category (optional, default: "all"): Feed category - "all", "hackernews", "tech", "ai", "research", or "community"
• sources (optional): Array of specific feed sources (overrides category)

Example Response:

Found 3 AI-related items from 6 source(s):

1. What could possibly go wrong if an enterprise replaces all its engineers with AI?
   Source: venturebeat
   https://venturebeat.com/ai/what-could-possibly-go-wrong...
   Published: Sat, 08 Nov 2025 05:00:00 GMT

2. Terminal-Bench 2.0 launches alongside Harbor
   Source: venturebeat
   https://venturebeat.com/ai/terminal-bench-2-0...
   Published: Fri, 07 Nov 2025 23:25:00 GMT

2. search_news

Search for news items matching specific keywords across multiple sources.

Parameters:

• keywords (required): Keywords to search for (comma-separated)
• limit (optional, default: 10): Maximum number of items to return
• category (optional, default: "all"): Feed category to search
• sources (optional): Array of specific feed sources to search

Example:

Keywords: "gpt-4, chatgpt"
Category: "ai"

3. list_sources

List all available news sources and categories.

No parameters required - returns complete list of 20 sources organized by category.

4. configure_schedule

Set up scheduled digest delivery.

Parameters:

• enabled (boolean): Enable/disable scheduler
• cronExpression (string): When to send (e.g., "0 8 * * *" for 8 AM daily)
• timezone (string): Your timezone (e.g., "Australia/Sydney")
• category (string): News category for digest
• limit (number): Number of items in digest

Example:

Configure schedule: enabled true, 8 AM daily, Australia/Sydney timezone

5. configure_notifications

Configure email and desktop notification settings.

Parameters:

• desktop (object): Desktop notification settings
• email (object): Email notification settings

Example:

Enable desktop notifications
Configure email to user@example.com

6. send_digest_now

Send a digest immediately (testing or on-demand).

Parameters:

• desktop (boolean): Send desktop notification
• email (boolean): Send email notification

Example:

Send me a digest now
Send digest via email only

7. get_config

View current configuration including schedule and notification settings.

Example:

Show my notification settings
What's my schedule configuration?

Testing

Method 1: Testing with Claude Desktop

Step 1: Configure Claude Desktop

Add the server to your Claude Desktop configuration file:

macOS Location:

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

Windows Location:

%APPDATA%/Claude/claude_desktop_config.json

Configuration:

{
  "mcpServers": {
    "ai-news": {
      "command": "node",
      "args": ["/absolute/path/to/mcp-ai-news-server/build/index.js"]
    }
  }
}

Important: Replace /absolute/path/to/mcp-ai-news-server with your actual project path.

To get your absolute path, run:

pwd

Step 2: Restart Claude Desktop

Completely quit and restart Claude Desktop for the changes to take effect.

Step 3: Verify Installation

In Claude Desktop, look for the hammer icon (🔨) in the input area, which indicates MCP tools are available.

Step 4: Test the Tools

Try these prompts in Claude Desktop:

Basic queries:

Get me the latest AI news
List all available news sources
Show me AI news from tech publications

Category-specific:

Get AI news from research sources
Show me news from AI company blogs
What's new on Hacker News about AI?

Advanced searches:

Search for "GPT-4" across all sources
Find news about "ChatGPT" in tech publications
Look for "transformer" in research papers
Get 15 AI news items from TechCrunch and VentureBeat

---

Method 2: Testing with MCP Inspector

The MCP Inspector provides a debugging interface for testing MCP servers.

Install MCP Inspector:

npm install -g @modelcontextprotocol/inspector

Run Inspector:

npx @modelcontextprotocol/inspector node build/index.js

This will open a web interface where you can:

• View available tools
• Test tool calls with custom parameters
• See request/response logs
• Debug server behavior

---

Method 3: Testing Standalone (Manual Testing)

You can test the server directly using Node.js with JSON-RPC messages.

Create a test script test.js:

import { spawn } from 'child_process';

const server = spawn('node', ['build/index.js']);

// Send initialize request
const initRequest = {
  jsonrpc: '2.0',
  id: 1,
  method: 'initialize',
  params: {
    protocolVersion: '2024-11-05',
    capabilities: {},
    clientInfo: { name: 'test-client', version: '1.0.0' }
  }
};

server.stdin.write(JSON.stringify(initRequest) + '\n');

// Send list tools request
setTimeout(() => {
  const listToolsRequest = {
    jsonrpc: '2.0',
    id: 2,
    method: 'tools/list',
    params: {}
  };
  server.stdin.write(JSON.stringify(listToolsRequest) + '\n');
}, 1000);

// Send call tool request
setTimeout(() => {
  const callToolRequest = {
    jsonrpc: '2.0',
    id: 3,
    method: 'tools/call',
    params: {
      name: 'get_ai_news',
      arguments: { limit: 5, source: 'newest' }
    }
  };
  server.stdin.write(JSON.stringify(callToolRequest) + '\n');
}, 2000);

server.stdout.on('data', (data) => {
  console.log('Response:', data.toString());
});

server.stderr.on('data', (data) => {
  console.error('Server log:', data.toString());
});

setTimeout(() => {
  server.kill();
}, 5000);

Run the test:

node test.js

---

Method 4: Testing with curl (JSON-RPC over stdio)

For quick testing, you can send JSON-RPC messages directly:

# Start the server
node build/index.js

# In another terminal, send a test message (note: this is complex for stdio)
echo '{"jsonrpc":"2.0","id":1,"method":"tools/list","params":{}}' | node build/index.js

---

Development

Project Structure

mcp-ai-news-server/
├── src/
│   └── index.ts          # Main server implementation
├── build/                # Compiled JavaScript (generated)
├── package.json          # Dependencies and scripts
├── tsconfig.json         # TypeScript configuration
└── README.md            # This file

Development Mode

# Watch mode (rebuild on changes)
npm run dev

Scripts

• npm run build - Compile TypeScript to JavaScript
• npm run start - Run the compiled server
• npm run dev - Build and run in one command

AI Keywords

The server filters content based on these keywords:

• ai, artificial intelligence
• machine learning, ml
• deep learning
• neural network
• llm, gpt, chatgpt
• claude, gemini
• openai, anthropic
• generative
• transformer
• nlp
• computer vision
• reinforcement learning

Troubleshooting

Server not appearing in Claude Desktop

1. Check config file path: Ensure you're editing the correct claude_desktop_config.json
2. Verify absolute path: The path in the config must be absolute, not relative
3. Check build: Run npm run build to ensure the server is compiled
4. Restart Claude Desktop: Completely quit and reopen the application
5. Check logs: Look at Claude Desktop's logs for errors

Finding Claude Desktop Logs

macOS:

~/Library/Logs/Claude/

Windows:

%APPDATA%/Claude/logs/

Testing if server starts correctly

node build/index.js
# Should output: "AI News MCP Server running on stdio"
# Press Ctrl+C to exit

RSS Feed Issues

If the RSS feeds are slow or unavailable:

• Check your internet connection
• The Hacker News RSS feeds might be temporarily down
• Try again in a few minutes

Environment

• Node.js: v20+ recommended
• TypeScript: v5+
• MCP SDK: v1.0.0+

License

MIT

Contributing

Contributions welcome! Please feel free to submit issues or pull requests.

Useful Links

• MCP Documentation
• MCP SDK
• Claude Desktop
• Hacker News RSS