RSSdeck-MCP

An MCP (Model Context Protocol) server that transforms RSS feeds into token-efficient, AI-ready data for agents like OpenClaw.

Overview

RSSdeck-MCP is designed for AI agents that need efficient, structured access to RSS content. Instead of fetching full articles (wasting tokens), it provides:

TL;DR Summaries — Concise 3-bullet summaries, not full text
Deduplication — Removes duplicate stories across feeds
Relevance Scoring — Filters by your interests
Sentiment Analysis — Bullish/bearish/neutral classification

Architecture

┌─────────────────┐     ┌──────────────────┐     ┌─────────────────┐
│   RSS Sources   │────▶│   RSSdeck-MCP    │────▶│  AI Agent       │
│   (30+ feeds)   │     │  (This Server)   │     │  (OpenClaw)     │
└─────────────────┘     └──────────────────┘     └─────────────────┘
                                │
                                ▼
                        ┌──────────────────┐
                        │  Token-Efficient │
                        │  Summaries        │
                        └──────────────────┘

Relationship to RSSdeck

RSSdeck (the main project) is a human-facing RSS reader:

Web UI with multi-column TweetDeck-style layout
Local AI integration (Ollama) for summaries
Full article scraping
For humans who want to read RSS manually

RSSdeck-MCP is an agent-facing data service:

MCP server for AI agents
Token-efficient API responses
No UI — purely programmatic
For AI agents that need RSS data

They work together:

Add feeds in RSSdeck (web UI)
Export to OPML
RSSdeck-MCP reads the OPML and serves agents

Installation

# Clone the repository
git clone https://github.com/mephistophelesbits/rssdeck-mcp.git
cd rssdeck-mcp

# Create virtual environment
python3 -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

# Install dependencies
pip install -r requirements.txt

Configuration

Create a .env file:

# RSSdeck instance URL (optional, for future API integration)
RSSDECK_URL=http://localhost:3001

# Your interests - comma separated (used for relevance scoring)
INTERESTS=AI,operations,Malaysia,APAC,business,technology,management

# Path to OPML file (exports from RSSdeck)
RSSDECK_OPML=./feeds/opml/your-feeds.opml

Usage

Running the MCP Server

python server.py

The server runs on stdio, ready to receive MCP calls from agents.

Using with OpenClaw

Configure OpenClaw to use RSSdeck-MCP as an RSS data source. The MCP server provides these tools:

| Tool | Description | |------|-------------| | get_feeds | List all configured RSS feeds | | get_updates | Get latest articles with TL;DR summaries | | search | Search across cached articles | | get_summary | Get detailed summary of a specific article |

Example Output

{
  "count": 5,
  "articles": [
    {
      "id": "abc123",
      "title": "Karpathy's Claws: The New Layer on AI Agents",
      "source": "Hacker News",
      "published": "2026-02-21T10:30:00Z",
      "tldr": "Andrej Karpathy discusses 'Claws' - AI agents on personal hardware...",
      "sentiment": "bullish",
      "relevance": 0.8,
      "url": "https://news.ycombinator.com/item?id=..."
    }
  ],
  "token_tip": "These are TL;DR summaries, not full articles."
}

Setting Up Feeds

Method 1: RSSdeck Export (Recommended)

Open RSSdeck at http://localhost:3001
Add your feeds in the UI
Click the Export OPML button in the sidebar
Save the OPML file to feeds/opml/
Update RSSDECK_OPML in .env to point to it

Method 2: Manual OPML

Create your own OPML file:

<?xml version="1.0" encoding="UTF-8"?>
<opml version="2.0">
  <head>
    <title>My Feeds</title>
  </head>
  <body>
    <outline text="Tech" xmlUrl="https://hnrss.org/frontpage"/>
    <outline text="AI" xmlUrl="https://simonwillison.net/atom/everything/"/>
  </body>
</opml>

Features

Token Efficiency

Summaries are capped at ~150 characters
Full articles are never sent to the agent
Deduplication reduces redundant content by ~40%

Relevance Scoring

Articles are scored based on your configured interests:

0.0 = irrelevant
1.0 = highly relevant

Sentiment Analysis

Simple keyword-based classification:

bullish — positive tech/business news
bearish — negative news (failures, vulnerabilities)
neutral — factual news

Use Cases

Second Brain

Power your personal AI assistant with curated RSS intelligence:

Daily AI/Tech news summaries
Industry-specific updates
Competitor monitoring

Automation

Feed AI agents structured data for:

Content generation
Trend analysis
Research pipelines

Monitoring

Track specific topics across many sources:

Job market intelligence
Tech industry news
Competitor updates

Integration with RSSdeck

┌─────────────────────────────────────────────────────────────┐
│                        RSSdeck                                │
│  ┌─────────────┐    ┌──────────────┐    ┌─────────────┐   │
│  │  Add Feeds │───▶│  Export OPML  │───▶│  MCP Server │   │
│  │  (Web UI)  │    │  (Sidebar)    │    │  (This)     │   │
│  └─────────────┘    └──────────────┘    └─────────────┘   │
└─────────────────────────────────────────────────────────────┘

Tech Stack

Python 3.10+ — Server runtime
MCP — Model Context Protocol
httpx — Async HTTP client
feedparser — RSS/Atom parsing
BeautifulSoup — HTML parsing

Contributing

Contributions welcome! Please read the contributing guidelines first.

License

MIT License — see LICENSE file.

Related Projects

RSSdeck — Human-facing RSS reader with AI integration
OpenClaw — AI agent platform that uses RSSdeck-MCP

Acknowledgments

Inspired by Karpathy's "Claws" concept
Built to power my second brain with OpenClaw

MCP Servers