DECEMBER 2025 | v2.2

SafeGamer.ai

Technical Overview — Roblox parental safeguards platform providing multi-layered AI-powered protection for children's gaming safety

63M
Daily Roblox Users
5
Beta Families
4
Protection Layers
3
LLM Models
90%+
Cost Savings via Caching
🎯
TARGET AUDIENCE
Technically proficient parents of children under 16, with focus on ages 13 and under playing Roblox
Parents Roblox Under 16
📊
CURRENT STATUS
Beta development with 5 test families in production, evolving from file-based to scalable SaaS
Beta 5 Families SaaS Ready
💰
BUSINESS MODEL
Freemium SaaS with premium tiers planned, focusing on AI-powered friend analysis as core differentiator
Freemium Premium Tiers
🚨 The Problem
  • 63M daily active users on Roblox, mostly children
  • Parents have limited visibility into online interactions
  • Predators use friend requests to groom children
  • Existing parental controls are binary (on/off)
  • No tool analyzes friend quality or suspicious patterns
  • Off-platform contact detection is nearly impossible
✅ Our Solution
  • AI-powered analysis of every friend connection
  • Multi-model LLM architecture for cost-effective analysis
  • Trust Network for cross-family verification
  • 4-layer protection strategy (no single point of failure)
  • Real-time alerts for high-risk friends
  • Visual network maps showing 2nd-degree connections

Key Differentiators

🛡️
MULTI-LAYERED DEFENSE
4 complementary protection layers ensure no single point of failure. Bypass one layer, others still protect.
🤖
AI FRIEND ANALYSIS
Multi-LLM analysis (Groq → GPT-4o → Claude) provides sophisticated risk assessment at minimal cost.
🤝
TRUST NETWORK
Cross-family verification creates a neighborhood watch for the digital world. Verified parents see each other's verifications.
💚
RELATIONSHIP PRESERVING
"Trust but verify" approach preserves parent-child relationships while maintaining necessary oversight.

System Architecture

Modern cloud-native stack deployed on Vercel

🎮
ROBLOX API
Friends, Profiles
🤖
LLM APIS
Groq, OpenAI, Anthropic
☁️
CLOUDFLARE
Pages, DNS
→
▲
VERCEL
Edge Functions
→
⚛️
NEXT.JS
Dashboard
⚙️
N8N
Workflows
🐍
SCRIPTS
Monitoring
→
🐘
SUPABASE
PostgreSQL + Auth
💾
CACHE
JSON Files
🔔
ALERTS
Discord, Email

Data Flow

End-to-end journey from Roblox API to parent dashboard

🎮
ROBLOX API
Public friend data
🐍
PYTHON SCRIPTS
Daily scheduled
⚙️
N8N WORKFLOWS
Automation
📸
FRIEND SNAPSHOT
Current state
🔄
COMPARE
Detect changes
🤖
LLM ANALYSIS
Groq → GPT-4 → Claude
💾
ANALYSIS CACHE
30-day TTL
🐘
SUPABASE
PostgreSQL
⚛️
NEXT.JS DASHBOARD
Parent interface
🚨
ALERTS
Discord / Email

Dashboard Features

📊
OVERVIEW
Risk distribution charts, key metrics, and high-risk friend alerts at a glance
👥
FRIENDS ANALYSIS
Searchable, filterable table with risk scores, red flags, and LLM explanations
🕸️
NETWORK MAP
Interactive vis.js graph showing friend connections and 2nd-degree relationships
✅
TRUST NETWORK
Cross-family verification badges showing verified SafeGamer parents
🔔
ALERTS
High/medium priority notifications with recommended parental actions
⚙️
SETTINGS
Family configuration, notification preferences, and account management

Multi-Model AI Architecture

Cost-optimized tiered approach to friend analysis

⚡
Groq
~$0.02/req
Fast Cheap
→
🧠
GPT-4o-mini
~$0.10/req
Medium
→
🎯
Claude
Fallback
Complex

What We Analyze

📝
PROFILE BIOS
Detect off-platform contact attempts (Discord usernames, Snapchat handles, phone numbers)
🏷️
USERNAMES
Flag inappropriate content, age indicators, scam patterns, and concerning terminology
🕸️
FRIEND NETWORKS
Analyze 2nd-degree connections, detect shared suspicious friends across children
📈
ACTIVITY PATTERNS
Monitor new friend velocity, friend turnover rates, and sudden changes

Risk Severity Levels

SAFE
No action needed
LOW
Monitor casually
MEDIUM
Review with child
HIGH
Immediate conversation
CRITICAL
Consider blocking

Cost Optimization

💾
INTELLIGENT CACHING
90%+ reduction in API calls by caching analysis results with 30-day TTL
90%+ Savings
🔄
CHANGE DETECTION
Only re-analyze friends with profile changes, not entire friend lists
📦
BATCH PROCESSING
Analyze multiple friends per API request to reduce overhead

4-Layer Protection Strategy

Multi-layered defense with no single point of failure

01 ACCOUNT-LEVEL

Built-in Roblox platform protections

🔐 Two-Factor Auth 📌 Account PIN 👥 Friends Only 🔒 Account Restrictions 💰 Purchase Controls
02 NETWORK-LEVEL

Router and DNS-based filtering

🔥 Firewalla 📡 UniFi 🌐 NextDNS ⏰ Time Limits 🚫 VPN Detection
03 DEVICE-LEVEL

On-device monitoring software

📱 Qustodio 📊 Screen Time 📝 Activity Logs 📍 Location Tracking
04 HUMAN-LEVEL — THE SAFEGAMER LAYER

AI-enhanced parental oversight — our core product

🤖 AI Friend Analysis 🕸️ Network Visualization ✅ Trust Network 🔔 Smart Alerts 📋 Weekly Reviews 💬 Safety Conversations

Technology Stack

Modern cloud-native architecture on Vercel

LAYER TECHNOLOGY STATUS
Frontend (Dashboard) Next.js 14 + React + Tailwind CSS + shadcn/ui Production
Frontend (Landing) Static HTML on Cloudflare Pages Production
Hosting Vercel (Edge Functions) Production
Database PostgreSQL (Supabase Cloud) Production
Authentication Supabase Auth (GoTrue) Production
LLM Analysis Groq → GPT-4o-mini → Claude Production
Automation n8n Workflows Production
Monitoring Scripts Python + Task Scheduler Production
Backend API Next.js API Routes / FastAPI Production

Beta Test Families

👨‍👩‍👧‍👦
ALPHA
Brad & Jack
👨‍👩‍👧
BRAVO
Christian & Dany Claire
👨‍👧
CHARLIE
Graham
👩‍👧‍👦
DELTA
Nella & Neva
👨‍👩‍👦
ECHO
Westin

Development Roadmap

From beta to scalable SaaS platform

NOW — DECEMBER 2025
Beta Production
5 families in production, Next.js dashboard on Vercel, Supabase Cloud integration, n8n workflows, Trust Network operational
Q1 2026
SaaS Foundation
Commercial SaaS architecture, OAuth providers (Google, GitHub), email infrastructure, advanced onboarding
Q2 2026
Feature Expansion
Advanced analytics, Minecraft monitoring, mobile-optimized views, enhanced reporting
Q3 2026
Scale & Optimize
Performance optimization, ML-based risk classification, additional gaming platforms
Q4 2026+
Growth
Scale to 1000+ families, mobile app, enterprise features

Key Metrics

METRIC CURRENT TARGET (1 YEAR)
Families 5 100+
Children Monitored ~10 200+
Friends Analyzed ~500 10,000+
LLM Cost/Month ~$0 (free tiers) <$200
Uptime 99% 99.9%

Resources

📂
GITHUB
github.com/zanebarker-ops/safegamer-ai
🌐
LANDING PAGE
safegamer-landing.pages.dev
📚
DOCUMENTATION
.claude/ directory in repository
📝
BETA WAITLIST
safegamer-landing.pages.dev/teaser/