GrokGuard

• 

  GrokGuard Blocking Scam Posts and Collapsing it.

Inspiration

X.com faces scams, impersonators, and misinformation. GrokGuard uses agentic AI to protect users in real time.

What it does

Real-time X.com feed scanner using Grok AI. Multi-agent system (Prosecutor, Defender, Arbiter) debates each post to detect scams, impersonators, and misinformation. Collapses dangerous posts, labels misinformation, and provides a live dashboard.

How we built it

• Chrome Extension (Manifest V3) + Grok API
• TypeScript agent orchestration (custom, no frameworks)
• Express.js API server + Next.js dashboard
• Real-time DOM monitoring for feed analysis

Challenges we ran into

• False positives on legitimate accounts → fixed with multi-factor legitimacy scoring
• Real-time performance → hybrid approach: instant analyzer for feed, full debate for profiles
• Extension context invalidation → heartbeat checks and state recovery

Accomplishments that we're proud of

• Truly agentic: autonomous investigation + multi-turn debate
• Zero false positives on verified accounts
• Real-time feed protection with instant AI analysis

What we learned

• Agentic AI needs clear agent roles and debate structure
• Hybrid models: fast for feed, reasoning for deep analysis
• Source credibility is critical for fact-checking

What's next for GrokGuard

• Human-in-the-loop learning from feedback
• Supabase integration for persistent analytics
• Web search integration for advanced fact-checking
• Mobile browser support

Built With

• chrome-extension-api
• express.js
• grok-api
• indexeddb
• next.js
• react
• typescript
• x-api