MCP‑XRelay

Automated Cybersecurity & Hardware News Poster for X (Twitter)

Fetches the latest security and enterprise‑hardware headlines from multiple RSS feeds, rewrites them using a local OpenAI‑compatible LLM, and posts concise updates to X every 5 minutes—completely offline, API‑free, and rate‑limit‑resilient.

---

✨ Key Features

⚙️	Description
Async RSS	Non‑blocking RSS fetch + thread‑pool parsing for fast performance
Local LLM	Works with any local OpenAI‑compatible endpoint (LM Studio, Ollama, llama.cpp)
Three Voices	Rotating personas: Friendly Expert, Thought‑Leader, Witty Advisor
Smart Tags	Global #TechNews + category tags (security, hardware, general)
CTA Links	30 % chance to append configurable call‑to‑action links
Retries	Exponential back‑off on LLM calls & posting (tenacity)
Session Reuse	Single Playwright browser context reused for all posts
Deduplication	SQLite database prevents duplicate tweets

---

🖥 Prerequisites

Tool	Version (tested)	Purpose
Python	3.9+	Runtime
pip / venv	latest	Dependency isolation
Playwright	≥ 1.40	Browser automation
Local LLM	Any OpenAI‑API‑compatible server	LLM inference endpoint

Default LLM model: gemma-3-1b-it-qat. Adjust LLM_MODEL if needed.

---

🔧 Environment Preparation

1 — Clone the Repository

git clone https://github.com/thebeardbro/MCP-XRelay.git
cd MCP-XRelay

2 — Create & Activate a Virtual Environment (recommended)

python -m venv .venv
# macOS / Linux
source .venv/bin/activate
# Windows (PowerShell)
.\.venv\Scripts\Activate.ps1
# Windows (CMD)
.\.venv\Scripts\activate.bat

3 — Install Dependencies & Playwright Browsers

pip install -r requirements.txt
playwright install               # one-time browser download (~120 MB)

4 — Configure Environment Variables

Create a .env file in the project root with the following:

# RSS feed URLs (comma‑separated)
FEED_URLS=https://feeds.feedburner.com/TheHackersNews,https://threatpost.com/feed/,https://securityaffairs.co/wordpress/feed

# Local LLM settings\LM_API_URL=http://127.0.0.1:1234/v1/chat/completions
LLM_MODEL=gemma-3-1b-it-qat

# Tweet configuration
CHAR_LIMIT=280
CTA_CHANCE=30

# Playwright session file
AUTH_STATE=state.json

# Database file\mDB_PATH=cybersec_news.db

# Optional CTA links override
CTA_LINKS=https://example.com/contact/,https://example.com/services/

5 — Capture X (Twitter) Session

One-liner to open Chromium, log in to X, and save cookies to state.json:

python -c "from playwright.sync_api import sync_playwright; p=sync_playwright().start(); b=p.chromium.launch(headless=False); ctx=b.new_context(); page=ctx.new_page(); page.goto('https://x.com/login'); input('Log in, then press Enter…'); ctx.storage_state(path='state.json'); b.close(); p.stop()"

This step must be done once before running the bot.

6 — Launch the Bot

python bot.py   # or python mcp_xrelay.py if renamed

The bot will run continuously, fetching feeds, generating tweets, and posting every 5 minutes. Use Ctrl +C to stop.

---

⚙️ Runtime Configuration

Override any setting via environment variables or inline:

FEED_URLS=https://myfeed.com/rss CTA_CHANCE=0 python bot.py

Important Variables

Variable	Purpose
FEED_URLS	Comma‑separated list of RSS feed URLs
LM_API_URL	Local LLM endpoint URL
LLM_MODEL	Model name for LLM inference
CHAR_LIMIT	Maximum tweet length
CTA_CHANCE	Percent chance to include a CTA link
AUTH_STATE	Playwright session file (cookies)
DB_PATH	SQLite database path for tweet history

---

📦 Docker (Optional)

FROM python:3.11-slim
WORKDIR /app
COPY . /app
RUN pip install -r requirements.txt && playwright install
ENV FEED_URLS="$(FEED_URLS)" \
    LM_API_URL=http://host.docker.internal:1234/v1/chat/completions \
    LLM_MODEL=gemma-3-1b-it-qat
CMD ["python", "bot.py"]

Mount your .env and state.json as volumes.

---

🧪 Sample Output

Patch Tuesday drops 60+ fixes in MSMQ—patch now to stay safe. https://threatpost.com/... #TechNews #CyberSecurity #ThreatIntel

---

🤝 Contributing

1. Fork the repo & create a branch
2. Make changes & run pre-commit run --all-files
3. Submit a pull request with a descriptive title

---

☕ Support

If MCP‑XRelay helps you, please:

• ⭐ Star the repository
• 🐦 Share on X
• ☕ Buy me a coffee

---

📄 License

MIT License © 2025 @thebeardbro