Crypto Risk Analysis Agent

An agent that can automatically analyze risk of a crypto meme coin

Comment

Crypto Risk Analysis Agent

An AI-powered intelligent agent system that automatically analyzes Solana token holder distributions and provides real-time risk assessments to protect investors from rugpulls, scams, and high-risk tokens.

🎯 Inspiration

The cryptocurrency space has seen explosive growth, but with it comes significant risk. Countless investors lose money to rugpulls, scam tokens, and highly concentrated holdings where a few wallets control the majority of supply. We witnessed friends and community members fall victim to these schemes, losing hard-earned money to projects with suspicious holder distributions that could have been detected with proper analysis.

The problem: Manual analysis of token holder distributions is time-consuming, requires blockchain expertise, and is nearly impossible to do at scale. By the time most investors discover warning signs, it's too late.

Our vision: Create an autonomous AI agent that continuously monitors token holder concentrations, analyzes wallet behaviors, and provides instant, actionable risk assessments—protecting investors before they get scammed.

💡 What it does

Crypto Risk Analysis Agent is a comprehensive risk intelligence platform that combines AI reasoning with blockchain data analysis to evaluate Solana tokens:

Core Features:

  1. Autonomous Token Analysis

    • Fetches top holder distributions for any SPL token
    • Calculates concentration metrics (top 3, 5, 10 holders)
    • Identifies exchange wallets, liquidity pools, and suspicious addresses
    • Generates risk scores based on holder patterns

  2. AI-Powered Risk Assessment

    • Uses Claude Opus 4 to provide human-like reasoning about risk factors
    • Analyzes holder concentration, liquidity presence, and manipulation risks
    • Delivers clear investment recommendations with actionable insights
    • Explains complex blockchain data in plain English

  3. Multi-Source Wallet Intelligence

    • Helius API: Analyzes token holder accounts and ownership data
    • Moralis API: Deep-dives into individual wallet portfolios, NFT holdings, and activity patterns
    • Detects burner wallets, developer wallets, and established traders
    • Identifies red flags like single-token holders or minimal activity

  4. Real-Time Dashboard

    • Live monitoring of analyzed tokens and risk levels
    • Agent activity logs showing autonomous decision-making
    • Visual risk indicators (Critical/High/Medium/Low)
    • Historical analysis tracking and trend detection

  5. Model Context Protocol (MCP) Server

    • Exposes blockchain analysis tools as MCP-compatible APIs
    • Enables AI agents to autonomously query blockchain data
    • Supports integration with Claude Desktop, AI workflows, and custom agents
    • Docker-ready deployment for cloud hosting

Risk Detection Capabilities:

  • ⚠️ High Concentration Risk: Top 10 holders control >70% of supply
  • 🚨 Whale Dominance: Single wallets holding disproportionate amounts
  • 🔍 Suspicious Patterns: New wallets with single-token holdings
  • 💧 Liquidity Issues: Lack of exchange/pool addresses in top holders
  • 🎭 Burner Wallets: Wallets with minimal SOL balance and activity
  • ✅ Positive Signals: Established wallets, exchange presence, good distribution

🛠️ How we built it

Architecture

Frontend (Next.js + React + Tailwind CSS)

  • Modern, responsive dashboard with real-time updates
  • Server-side rendering for performance
  • Elegant UI with Lucide icons and custom components
  • Deployed and optimized for production

Backend (Node.js + Express)

  • RESTful API endpoints for token risk analysis
  • Integration with MCP client for tool orchestration
  • Real-time data processing and caching
  • Error handling and rate limiting

AI Layer (Anthropic Claude Opus 4)

  • Advanced reasoning about token holder distributions
  • Context-aware risk assessment with natural language explanations
  • Structured prompts for consistent, professional analysis
  • Integration via Anthropic SDK

MCP Server (Model Context Protocol)

  • Custom-built tools for blockchain data access:
    • find_account_info: Analyzes token holder concentration (Helius)
    • analyze_wallet_risk: Evaluates individual wallet risk profiles (Moralis)
    • get_wallet_portfolio: Fetches complete wallet holdings (Moralis)
    • search_tokens: Searches tokens on DexScreener
  • Supports HTTP, SSE, and stdio transports
  • Dockerized for easy deployment to cloud platforms

Data Layer (Sanity CMS)

  • Headless CMS for storing risk reports and analysis history
  • Real-time data sync with frontend
  • Structured schemas for coins, risk assessments, and agent logs
  • Content API for querying and filtering

Blockchain Integration

  • Helius RPC: Solana mainnet access for token account queries
  • Moralis API: Multi-chain wallet and portfolio data
  • Rate-limited API calls with retry logic
  • Efficient data parsing and transformation

Tech Stack

  • Frontend: Next.js 15, React 19, TypeScript, Tailwind CSS
  • Backend: Node.js, Express, ES6 modules
  • AI: Anthropic Claude Opus 4, Google Gemini, Model Context Protocol (MCP)
  • Blockchain: Helius API, Moralis API, Solana Web3.js, TRM Labs
  • Database: Sanity CMS (headless)
  • DevOps: Docker, DigitalOcean, bash deployment scripts
  • APIs & Tools: Postman, RESTful architecture, JSON-RPC 2.0

Integrations

  • Moralis API: Solana wallet analysis, portfolio tracking, NFT data
  • TRM Labs: Blockchain intelligence and compliance
  • Sanity CMS: Headless content management and data storage
  • Postman: API testing, documentation, and workflow automation
  • Anthropic Claude: AI-powered risk reasoning and analysis
  • Google Gemini: Additional AI model for enhanced insights

🚧 Challenges we ran into

1. Rate Limiting Nightmares

Helius API's getTokenLargestAccounts method triggered aggressive rate limiting when analyzing tokens with many holders (like BONK). We couldn't simply reduce the number of accounts—the API queries all accounts internally regardless.

Solution: Implemented sequential processing with 500ms delays between requests, error detection for rate limit responses, and graceful degradation. Added retry logic and fallback strategies for production reliability.

2. MCP Protocol Learning Curve

Model Context Protocol was new to all of us. Understanding tool schemas, transport layers (HTTP vs SSE vs stdio), and integration patterns took significant research and experimentation.

Solution: Built incrementally—started with simple tools, tested extensively with Postman, then integrated with Claude Desktop. Created comprehensive examples and documentation for future developers.

3. Docker Health Checks Failing

Initial Dockerfile didn't expose ports or run the server in HTTP mode, causing DigitalOcean to repeatedly kill containers due to failed health checks.

Solution: Rewrote Dockerfile to use --streamable-http flag by default, added EXPOSE 3001, installed curl for health checks, and implemented proper HEALTHCHECK directive.

4. AI Reasoning Quality

Getting Claude to provide consistent, actionable risk assessments (not generic advice) required extensive prompt engineering.

Solution: Developed structured prompts with specific sections (Holder Distribution, Red Flags, Liquidity, Market Manipulation, Recommendations). Provided concrete data points and metrics to ground the AI's reasoning in facts.

5. Multi-API Data Aggregation

Combining data from Helius (token accounts), Moralis (wallet portfolios), and DexScreener (token metadata) with different formats and rate limits was complex.

Solution: Created abstraction layers with consistent error handling, implemented parallel fetching where possible, and built robust parsing logic for various response structures.

🏆 Accomplishments that we're proud of

✨ Built a Production-Ready AI Agent System: Not just a prototype—fully deployed, containerized, and operational with real blockchain data.

🧠 Advanced AI Reasoning Integration: Successfully leveraged Claude Opus 4 to provide human-expert-level analysis of complex blockchain data that goes beyond simple metrics.

🔗 Model Context Protocol Implementation: One of the early adopters to build a complete MCP server with multiple blockchain analysis tools, complete with documentation and deployment scripts.

🎨 Beautiful, Intuitive Dashboard: Created a professional-grade UI that makes complex risk data accessible to non-technical users while providing depth for power users.

🚀 Real-World Impact: Analyzed actual tokens like BONK with 93+ trillion supply and 20+ holders, providing actionable insights that could prevent real investment losses.

📊 Multi-Source Intelligence: Integrated three major blockchain data providers (Helius, Moralis, DexScreener) into a cohesive analysis pipeline.

🐳 DevOps Excellence: Wrote automated deployment scripts, Dockerfiles with proper health checks, and environment variable management for seamless cloud deployment.

📚 What we learned

Technical Skills

  • Model Context Protocol (MCP): Deep understanding of tool schemas, transport layers, and AI agent integration patterns
  • Blockchain Data Analysis: Learned intricacies of SPL token accounts, holder distributions, and on-chain data structures
  • AI Prompt Engineering: Mastered techniques for getting consistent, high-quality reasoning from LLMs with structured prompts
  • Rate Limit Management: Strategies for working within API constraints while maintaining functionality
  • Docker Optimization: Health checks, multi-stage builds, and production-ready containerization

Product Insights

  • User Education is Key: Risk analysis is only valuable if users understand it—we need to explain WHY a token is risky, not just assign a score
  • Speed Matters: Users want instant assessments; we optimized for sub-5-second analysis despite multiple API calls
  • Trust Through Transparency: Showing raw data, methodology, and AI reasoning builds confidence in our assessments

AI Agent Development

  • Tool Design Philosophy: Good MCP tools should be single-purpose, well-documented, and return structured data
  • Context is Everything: LLMs produce better analysis when given rich context (holder types, concentration metrics, historical patterns)
  • Autonomous Decision Making: AI agents can make sophisticated investment risk decisions when given proper tools and reasoning frameworks

Blockchain Ecosystem

  • Data Fragmentation: No single API provides complete token intelligence—aggregation is essential
  • Solana's Complexity: SPL token architecture (mint addresses, token accounts, associated accounts) requires careful handling
  • DeFi Patterns: Learned to identify exchange wallets, liquidity pools, and common holding patterns that indicate legitimacy

🚀 What's next for Crypto Risk Analysis Agent

Short-Term Roadmap (Next 3 Months)

1. Multi-Chain Expansion

  • Add Ethereum, Base, and Polygon support
  • Cross-chain portfolio risk analysis
  • Unified risk scoring across networks

2. Advanced Risk Models

  • Machine learning models trained on historical rugpull data
  • Predictive analytics for emerging risks
  • Smart contract vulnerability scanning integration

3. Real-Time Alerts

  • Webhook notifications for high-risk token launches
  • Twitter bot for instant community warnings
  • Telegram/Discord integration for project alerts

4. Community Features

  • User-submitted token analyses
  • Reputation system for wallet addresses
  • Collaborative risk flagging

Medium-Term Vision (6-12 Months)

5. Autonomous Agent Swarm

  • Multiple specialized agents (Contract Auditor, Liquidity Hunter, Social Sentiment Analyzer)
  • Agent collaboration and consensus-building
  • Distributed analysis across token networks

6. Developer Platform

  • Public API for risk scores
  • Embeddable widgets for DEX platforms
  • SDK for integrating risk analysis into wallets

7. Portfolio Protection

  • Automated wallet monitoring
  • Pre-transaction risk checks (browser extension)
  • Smart wallet integration with auto-reject for high-risk tokens

8. Advanced Analytics

  • Historical trend analysis and pattern detection
  • Comparative analysis across similar tokens
  • Risk forecasting and probability models

Long-Term Goals (12+ Months)

9. Regulatory Compliance Tools

  • KYC/AML integration for institutional use
  • Compliance reporting and audit trails
  • Regulatory risk indicators

10. Insurance Integration

  • Partner with DeFi insurance protocols
  • Risk-based premium calculations
  • Automated claim filing for rugpulls

11. Decentralized Intelligence Network

  • On-chain risk oracle for smart contracts
  • Token-gated access to premium analysis
  • DAO governance for risk methodology

12. AI Model Marketplace

  • Allow community to train and contribute risk models
  • Competing AI agents with performance tracking
  • Incentivize high-accuracy predictions

🔗 Links

👥 Team

Built with passion during the hackathon by developers committed to making crypto safer for everyone.

⚠️ Disclaimer: This tool provides risk analysis for informational purposes only. Always do your own research (DYOR) before investing. Not financial advice.

Built With

Share this project:

Updates