What it does OpenNews MCP is a thin Python wrapper around the 6551.io API. It exposes news, market data, and on-chain signals as tools that any MCP-compatible AI assistant (Claude, OpenClaw) can call. You ask for “bullish BTC signals above impact score 80,” it fetches and filters. Real-time updates run over a WebSocket.

The interesting bit The README includes a pre-packaged security audit prompt: paste it into your AI assistant and it reviews the source code for hardcoded tokens, suspicious dependencies, and data exfiltration before you install. It’s a neat admission that “trust me, it’s safe” doesn’t cut it anymore.

Key highlights

• 84+ sources across six categories: news (Bloomberg, Reuters, TASS, Truth Social), exchange listings, on-chain whale trades, meme-coin sentiment, market anomalies, and 12 AI prediction signals
• Every article gets an AI impact score (0–100), a trading signal (long/short/neutral), and bilingual EN/ZH summaries
• WebSocket live subscriptions with coin and source filtering
• Token-gated: requires free API key from 6551.io/mcp
• Supports Claude Code one-liner install and OpenClaw skill copy

Caveats

• The project itself is mostly glue: API client, config loader, and tool definitions. The heavy lifting (aggregation, AI scoring, WebSocket backend) happens at 6551.io, not in this repo
• “AI-powered prediction” signals like CORRELATION_CAUSAL and SETTLEMENT_ARBITRAGE are named but not explained; what they actually compute is unclear
• Source list includes duplicates (“Handelsblatt” and “Hadelsblatt”) and some oddities (Ambrey maritime intelligence next to Politico)

Verdict Worth a look if you want your AI assistant to monitor crypto markets without building scrapers yourself. Skip it if you need transparent, auditable signal logic or are allergic to third-party API dependencies.