Protect MCP – Instant protection from AI prompt injection attacks.

AI agents calling external web data (LinkedIn scrapers, MCP integrations) often fall victim to prompt injections. Malicious websites can trick agents into illegal actions or leaking data.

That can happen to anyone:

What it does

Protect MCP is an open-source security middleware that instantly intercepts, detects, and blocks malicious prompt injections on MCP calls and external web requests. Alerts go directly to WhatsApp, Email, or MixPanel.

How we built it • FastAPI backend for instant interception & validation. • One-click CLI to deploy to vercel with instant protection and custom policies • SDK with simple proxy setup for secure integration. • MixPanel integration for real-time security analytics.

Built, launched, and open-sourced during A2A hackathon.