Behavioural Prediction MCP Server: Wallet Fraud Detection

Overview

The Behavioural Prediction MCP Server from ChainAware.ai provides AI-driven risk scoring and behavioural analysis for blockchain wallets and contracts. It exposes MCP-compatible tools that help developers, exchanges, and DeFi platforms detect fraud, rug pulls, sanctions exposure, and other malicious activity before funds are at risk.

The server implements the Model Context Protocol (MCP) and offers a Server‑Sent Events (SSE) endpoint for streaming responses. Access is controlled via API key (available by request). Typical integrations use the predictive fraud and behavioural analysis tools to add automated risk gating, alerts, or trust scoring into transaction flows and monitoring pipelines.

Features

• AI-powered wallet risk scoring (fraud / AML indicators)
• Rug-pull and behavioural-risk forecasting for contracts and liquidity events
• Forensic detail breakdown (phishing, mixers, sanctioned exposure, darkweb links, etc.)
• MCP-compliant toolset for easy integration with MCP clients (including Claude/other MCP consumers)
• SSE streaming endpoint for real-time responses
• Multi-chain support (ETH, BNB, POLYGON, TON, BASE, TRON, HAQQ, etc.)
• Access control via API keys; audit fields (createdAt / updatedAt) included in responses

Installation / Configuration

This server is hosted by ChainAware and is consumed — you do not install the server locally. To integrate:

1. Request API access from ChainAware.ai and obtain an API key.
2. Call the MCP SSE endpoint: https://prediction.mcp.chainaware.ai/sse
3. Supply your API key and tool-specific inputs in the request body or headers as directed by your MCP client.

Example: curl (SSE request)

curl -N \
  -H "Accept: text/event-stream" \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "tool": "predictive_fraud",
    "input": { "network": "ETH", "walletAddress": "0x1234...abcd" }
  }' \
  https://prediction.mcp.chainaware.ai/sse

Example: JavaScript (fetch, basic POST to SSE with fetch; use EventSource or an MCP client for full SSE handling)

const payload = {
  tool: "predictive_fraud",
  input: { network: "ETH", walletAddress: "0x1234...abcd" }
};

fetch("https://prediction.mcp.chainaware.ai/sse", {
  method: "POST",
  headers: {
    "Content-Type": "application/json",
    "Authorization": "Bearer YOUR_API_KEY",
    "Accept": "text/event-stream"
  },
  body: JSON.stringify(payload)
}).then(res => {
  // handle server-sent events in browser with EventSource or a streaming library
  console.log("Connected; use SSE client to process events.");
});

Available Tools

The MCP server ships multiple tools (each exposed as an MCP tool id). The primary ones:

• predictive_fraud

  • Purpose: Forecasts likelihood of fraudulent activity for a wallet, performs AML-style checks, and returns forensic indicators.
  • Typical output: classification (Fraud / Not Fraud / New Address), probability score, forensic_details, sanctionData, timestamps.

• predictive_behaviour

  • Purpose: Projects wallet or contract behavioural patterns and liquidity risk, used to predict rug-pulls and abnormal transaction sequences. Returns risk scores, signal timelines, and recommendations for monitoring thresholds.
  • Typical output: behaviour classification, probability of rug-pull or abnormal exit, sequence of flagged events, recommended actions.

Note: The MCP tool identifiers are:

• io.github.ChainAware/chainaware-behavioral-prediction-mcp (server mcp-name)

Refer to the repository for any updates: https://github.com/ChainAware/behavioral-prediction-mcp

predictive_fraud — API Reference (summary)

Inputs

Example JSON response (abridged)

{
  "message": "Success",
  "walletAddress": "0x1234...abcd",
  "chain": "ETH",
  "status": "Not Fraud",
  "probabilityFraud": "0.02",
  "forensic_details": {
    "phishing_activities": "low",
    "mixer": "none",
    "sanctioned": "false"
  },
  "checked_times": 1,
  "createdAt": "2026-01-12T10:00:00Z",
  "updatedAt": "2026-02-01T15:22:30Z",
  "sanctionData": []
}

Error cases

• 403 Unauthorized — invalid or missing API key
• 400 Bad Request — malformed network or walletAddress
• 500 Internal Server Error — temporary downstream failure

Use Cases

• Exchange deposit gating: Block or flag incoming deposits from high-probability fraud wallets before crediting user balances.
• Smart-contract monitoring: Combine predictive_behaviour outputs with on-chain event listeners to detect early signs of rug-pull patterns (sudden liquidity removal, dev wallet sweeps).
• AML workflows: Enrich compliance pipelines with forensic details (mixer interactions, darkweb links, sanction lists) to prioritize investigations.
• Wallet onboarding: Score new addresses to decide whether to allow certain privileged operations (token listings, airdrops, governance participation).
• Analytics dashboards: Surface risk metrics per wallet and aggregate risk trends by contract or token.

Resources

• Repository: https://github.com/ChainAware/behavioral-prediction-mcp
• Server SSE URL: https://prediction.mcp.chainaware.ai/sse
• Company: https://chainaware.ai/
• MCP spec: https://github.com/modelcontextprotocol/spec
• Twitter: https://x.com/ChainAware

For production integrations use an MCP client that supports SSE and tool selection. Contact ChainAware.ai to request API access and production keys.