Attestr is a multi-agent DeFi research and smart contract analysis platform. It combines live web search, on-chain data, and AI synthesis into actionable insights — accessible via Claude.ai, Claude Code, and the CROO Agent Store.
Live URLs
| Service | URL |
|---|---|
| Frontend | https://attestr-frontend.vercel.app |
| MCP Server | https://attestr-mcp-production.up.railway.app |
| Backend (CAP) | Railway — 24/7 |
Attestr exposes four intelligence services, paid in USDC via the CROO Agent Protocol (CAP):
| Service | What it does | Pricing |
|---|---|---|
| Web3 Research | Live Google search + source verification + AI synthesis → structured report | Per query, in USDC |
| Contract Risk | On-chain analysis via Etherscan (Base) → SAFE / CAUTION / DANGEROUS badge + score/100 | Per address, in USDC |
| Due Diligence | Research + risk in parallel → combined markdown report + overall confidence score | Per query, in USDC |
| Hyperliquid Vault | Vault TVL, APR, commission, capacity → YES/NO deposit recommendation + risk score | Per vault, in USDC |
Prices are configured in the CROO Dashboard per service. All payments settle on Base mainnet in USDC.
- Go to Claude.ai → Settings → Integrations → Add MCP Server
- Enter:
https://attestr-mcp-production.up.railway.app/mcp(must include the/mcppath — Claude.ai sends protocol calls to exactly the URL you enter, not a discovered sub-path) - Claude.ai will redirect you to the OAuth authorization page
- Enter your
MCP_API_KEYto grant access - All 6 tools are now available in Claude.ai
claude mcp add attestr --transport http https://attestr-mcp-production.up.railway.app/mcpSet the Authorization header using your MCP_API_KEY:
{
"mcpServers": {
"attestr": {
"type": "http",
"url": "https://attestr-mcp-production.up.railway.app/mcp",
"headers": {
"Authorization": "Bearer YOUR_MCP_API_KEY"
}
}
}
}git clone https://github.com/mdfaizee19/ATTESTR-CROO.git
cd ATTESTR-CROO
cp .env.example .env # fill in your API keys
npm install
npm run mcp:http # starts HTTP MCP server on port 3001Check if a contract is safe:
check_contract_risk("0x833589fCD6eDb6E08f4c7C32D4f71b54bdA02913")
→ { badge: "SAFE", riskScore: 10, reasons: [...], report: "..." }
Research a DeFi topic:
research_web3("What are the risks of Uniswap V4 hooks?")
→ { executiveSummary, keyFindings, riskAssessment, confidenceScore: 82 }
Analyze a Hyperliquid vault:
analyze_hyperliquid_vault("0x010461C14e146ac35Fe42271BDC1134EE31C703a")
→ { tvlUsd: 420000, apr: 18.5, badge: "SAFE", depositRecommendation: "YES" }
Full due diligence:
full_due_diligence("Is Aave V3 safe on Base? 0xA238Dd80C259a72e81d7e4664a9801593F98d1c5")
→ Combined report: research + on-chain risk + overall confidence score
Check if the agent is online:
get_agent_status()
→ { status: "online", uptime: "2h 14m 3s", activeOrders: 0, services: [...] }
List incoming orders:
list_orders(status: "paid")
→ { total: 3, orders: [{ orderId, serviceName, status, createdAt, price }, ...] }
| Layer | Technology |
|---|---|
| Agent protocol | CROO Agent Protocol (CAP) over Base mainnet |
| Payment | USDC on Base (0x833589fCD6eDb6E08f4c7C32D4f71b54bdA02913) |
| MCP transport | StreamableHTTP (RFC 8414 OAuth 2.0) |
| AI synthesis | Groq — llama-3.3-70b-versatile |
| Web search | Serper (Google Search API) |
| On-chain data | Etherscan v2 API (Base, chainId 8453) |
| Vault data | Hyperliquid REST API |
| Backend runtime | Node.js + TypeScript (ts-node) |
| Frontend | React + Vite, deployed on Vercel |
| Backend hosting | Railway (24/7) |
┌─────────────────────────────────────────────────────────────┐
│ Users │
│ Claude.ai · Claude Code · CROO Agent Store │
└────────────┬──────────────────────────────┬─────────────────┘
│ MCP (HTTP + OAuth) │ CAP (WebSocket + USDC)
▼ ▼
┌─────────────────────┐ ┌─────────────────────────────┐
│ MCP Server │ │ Coordinator │
│ /mcp (HTTP) │ │ src/coordinator/index.ts │
│ Railway │ │ Railway │
└────────┬────────────┘ └──────────────┬──────────────┘
│ │
│ Groq · Serper · Etherscan │ Groq · Serper
│ Hyperliquid │ Etherscan · Hyperliquid
└──────────────┬─────────────────────┘
│
┌─────────▼──────────┐
│ External APIs │
│ Groq Serper │
│ Etherscan HL │
└────────────────────┘
MCP path: Claude → OAuth → /mcp → tools run inline → response
CAP path: User pays USDC on Base → NegotiationCreated → OrderPaid → pipeline runs → deliverOrder
Attestr/
├── src/
│ ├── coordinator/ # CAP backend — CROO WebSocket agent
│ │ └── index.ts
│ ├── mcp/ # MCP HTTP server — Claude.ai connector
│ │ └── index.ts
│ ├── config.ts # CROO network config + requireEnv
│ └── types.ts # Shared TypeScript interfaces
├── frontend/ # React landing page (Vercel)
├── docs/ # Integration guides
│ ├── CAP-INTEGRATION.md
│ └── MCP-INTEGRATION.md
├── .env.example # Environment variable template
└── package.json
Copy .env.example to .env:
# AI providers
GROQ_API_KEY= # Groq — AI synthesis
SERPER_API_KEY= # Serper — web search
ETHERSCAN_API_KEY= # Etherscan v2 — on-chain data
# MCP server
MCP_API_KEY= # Bearer token for Claude.ai OAuth
SERVER_URL= # Public URL (Railway or ngrok)
PORT=3001
COORDINATOR_URL= # Coordinator's internal URL, used by get_agent_status (e.g. http://attestrcroo.railway.internal:8080)
DASHBOARD_SECRET= # Shared secret between MCP server and coordinator's /health endpoint
# CROO (CAP backend + list_orders tool)
CROO_SDK_KEY_COORDINATOR=
CROO_SERVICE_ID_RESEARCH=
CROO_SERVICE_ID_RISK_CHECK=
CROO_SERVICE_ID_DUE_DILIGENCE=
CROO_SERVICE_ID_HYPERLIQUID=See MCP-INTEGRATION.md for MCP setup and CAP-INTEGRATION.md for CAP/CROO integration.