Revealing What RWA Projects Won't Disclose
The AI-powered Trust Layer for Real World Assets -- bringing S&P-grade transparency to crypto's $323B stablecoin market and beyond.
Stablecoins are crypto's first killer product -- $323B market cap, $46T annual volume (surpassing Visa + Mastercard combined). The US passed the GENIUS Act recognizing stablecoins as strategic dollar infrastructure.
But here's the dirty secret: the trust foundation of every RWA token is an audit report that almost nobody reads.
- The CFTC found Tether held adequate reserves for only 27.6% of a 26-month period ($41M fine)
- It took Tether 10 years to obtain a Big Four audit
- When retail investors notice a de-peg, the bank run is already underway -- institutions with dedicated analysts always move first
RWA has an Oracle Problem: the bridge between off-chain assets and on-chain tokens depends on audit reports that are complex, infrequent, and favor institutions over retail.
RWANDA is an autonomous AI Agent system that continuously monitors, analyzes, and grades RWA projects -- making institutional-grade due diligence accessible to everyone.
Think of it as the S&P of Real World Assets, but AI-native and on-chain.
Fetch --> Analyze --> Record --> Alert
- Fetch -- AI Agent periodically crawls audit reports, attestations, and on-chain data from RWA issuers
- Analyze -- Category-specific evaluation frameworks score each project across multiple dimensions
- Record -- Results are written to an on-chain registry (Solidity smart contract) for immutability
- Alert -- Dashboard updates in real-time; Telegram bot pushes instant notifications on grade changes
Different RWA categories require fundamentally different evaluation criteria:
Stablecoins (USDT, USDC):
- GENIUS Act Compliance -- Does the reserve meet new US regulatory standards?
- Reserve Adequacy -- Total assets vs. total liabilities ratio
- Reserve Composition -- % in T-Bills vs. risky assets (crypto, precious metals, secured loans)
- Custody & Jurisdiction -- Where are assets held? Under what legal framework?
- Reporting & Audit Quality -- Big Four? ISAE 3000R? How frequently?
Tokenized Equities (Ondo GM):
- Collateral Ratio -- Per-token backing verification
- Verification Frequency -- Daily attestation vs. quarterly reports
- Bankruptcy Remoteness -- Is the SPV structure legally isolated?
- Custody & Counterparty -- Who holds the underlying securities?
- Reporting Freshness -- How stale is the latest verification?
Each dimension produces a weighted score, aggregated into a letter grade (A+ to F) that anyone can understand.
RWANDA is deployed and functional:
- $270B+ in RWA assets monitored
- 267 tokens tracked across stablecoin and tokenized equity categories
- 277 audit reports analyzed
- 3 projects with full analysis: Tether USDT (Grade C), Circle USDC (Grade A), Ondo GM (Grade B+)
- Automated cron jobs running on Vercel (daily/monthly/quarterly per project)
- Telegram alert bot for real-time grade change notifications
- On-chain registry smart contract for immutable trust records
rwa-nda/
+-- app/ # Next.js 16 App Router
| +-- page.tsx # Dashboard -- project overview + trust scores
| +-- project/[id]/ # Deep-dive analysis per project
| +-- api/cron/analyze/ # Automated analysis pipeline (Vercel Cron)
+-- components/ # React components
| +-- ProjectTable.tsx # Main project listing with grades
| +-- TrustScoreRadar.tsx # Radar chart (Chart.js) for dimensions
| +-- ReserveComposition # Reserve breakdown visualization
| +-- RedFlagsList.tsx # Severity-coded red flag display
| +-- AlertCTA.tsx # Telegram subscription CTA
+-- contracts/
| +-- RWANDARegistry.sol # On-chain trust score registry (Solidity 0.8.20)
+-- data/
| +-- projects.json # Project metadata
| +-- analyses/ # Analysis results (JSON)
+-- lib/
| +-- analyzer.ts # Core analysis engine + grading logic
| +-- types.ts # TypeScript type definitions
| +-- supabase.ts # Database integration
| +-- telegram.ts # Telegram alert system
| +-- ondo.ts # Ondo token list fetcher
+-- telegram-bot/
| +-- bot.js # Telegram bot (subscribe/check/alerts)
| +-- send-alert.js # Push alert dispatcher
| Layer | Technology |
|---|---|
| Frontend | Next.js 16, React 19, Tailwind CSS 4 |
| AI Engine | Anthropic Claude API (analysis pipeline) |
| Database | Supabase (PostgreSQL) |
| On-Chain | Solidity 0.8.20 (trust record registry) |
| Automation | Vercel Cron Jobs |
| Alerts | Telegram Bot API |
| Charts | Chart.js + react-chartjs-2 |
| Deployment | Vercel |
- Node.js 18+
- npm or yarn
git clone https://github.com/your-repo/rwa-nda.git
cd rwa-nda
npm installCopy .env.example and fill in your keys:
cp .env.example .env| Variable | Description |
|---|---|
NEXT_PUBLIC_SUPABASE_URL |
Supabase project URL |
SUPABASE_SERVICE_ROLE_KEY |
Supabase service role key |
CRON_SECRET |
Secret for Vercel Cron authentication |
TELEGRAM_BOT_TOKEN |
Telegram bot token from @BotFather |
npm run devOpen http://localhost:3000 to see the dashboard.
cd telegram-bot
cp .env.example .env # Add your TELEGRAM_BOT_TOKEN
npm install
npm startCommands: /start, /subscribe, /check <project>, /list
Automated analysis runs on Vercel Cron:
| Project | Schedule | Endpoint |
|---|---|---|
| Ondo GM | Daily | /api/cron/analyze?project=ondo-gm |
| Circle USDC | Monthly | /api/cron/analyze?project=circle-usdc |
| Tether USDT | Quarterly | /api/cron/analyze?project=tether-usdt |
Schedules align with each project's actual reporting frequency.
Phase 1: Trust Layer (Now) -- The first place retail investors check before trusting an RWA token.
Phase 2: De Facto Standard -- RWA projects seek RWANDA certification; exchanges require RWANDA grades for listing.
Phase 3: RWA Infrastructure -- S&P stopped at ratings. RWANDA leverages its trust layer to become the issuance infrastructure itself -- every RWA issued through RWANDA.
This project was built during the SEABW 2026 "Play to Build" AI Hackathon (24-hour vibe coding challenge). The entire codebase -- from smart contracts to the analysis engine to the frontend -- was developed in collaboration with AI coding tools. Commit history reflects AI-assisted development throughout.
MIT