INITIALIZATION

Go to Claude, and under settings, choose Connectors. Add a new connector and paste the following link: "https://fraud-alert-mcp-ha-hak-team-amrita-university-amritapuri-campus.app.nitrocloud.ai/mcp" Give it the required action permissions and reference it by a name. You can now use this in Claude.

Fraud Alert MCP Server

A Model Context Protocol (MCP) server for real-time fraud detection and transaction analysis. Integrates with Claude and other AI models to provide intelligent fraud risk assessment.

What is MCP?

The Model Context Protocol (MCP) is a standard that allows AI models like Claude to safely invoke tools and access resources from external servers. This MCP server exposes fraud detection tools that Claude can call to analyze transactions, flag suspicious activity, and manage fraud rules.

How to Use This Server

1. Local Development

# Install dependencies
npm install

# Start the MCP server
npm run dev

The server will start and be ready to accept MCP connections.

2. Connect in Claude Desktop

Add this server to your claude_desktop_config.json:

{
  "mcpServers": {
    "fraud-alert": {
      "command": "node",
      "args": ["dist/index.js"]
    }
  }
}

Then restart Claude Desktop. The fraud detection tools will appear in the Tools panel.

3. Programmatic Integration

Use any MCP client library to connect:

// Example: Node.js MCP client
const client = new MCPClient();
await client.connect('fraud-alert');
const result = await client.callTool('analyze_transaction', {
  transactionId: 'TX-123',
  userId: 'U-456',
  amount: 5000,
  // ... other fields
});

Available Tools

analyze_transaction

Analyzes a transaction for fraud risk. Returns a risk score (0-100), identified risk factors, and a recommended action.

Input:

• transactionId (string): Unique transaction ID
• userId (string): User ID
• amount (number): Transaction amount
• currency (string): Currency code (e.g., USD)
• merchantCategory (string): Merchant category (e.g., electronics)
• location (object): Transaction location with lat/long/country
• deviceId (string): Device identifier
• timestamp (string): ISO 8601 timestamp

Output:

• isFraudulent (boolean): Fraud flag
• riskScore (0-100): Risk score
• riskFactors (array): List of identified risk factors
• recommendedAction (enum): ALLOW, REVIEW, or BLOCK

flag_transaction_status

Updates the status of a transaction and triggers an alert.

Input:

• transactionId (string): Transaction ID
• actionTaken (enum): BLOCKED, FLAGGED_FOR_REVIEW, or CLEARED
• agentReasoning (string): Reason for the action

Output:

• success (boolean): Operation status
• alertId (string): Unique alert ID
• timestamp (string): Alert timestamp

Project Structure

fraud-alert-mcp/
├── src/
│   ├── modules/
│   │   ├── detection/          # Fraud detection logic
│   │   │   ├── detection.tools.ts
│   │   │   ├── detection.service.ts
│   │   │   └── detection.module.ts
│   │   ├── calculator/         # Calculator tools
│   │   ├── governance/         # Governance tools
│   │   └── ...
│   ├── widgets/                # UI widgets for tool results
│   ├── app.module.ts           # Main app module
│   └── index.ts                # Server entry point
├── package.json
├── tsconfig.json
└── README.md

Environment Variables

Create a .env file:

ANTHROPIC_API_KEY=sk-ant-...
PORT=3000

Testing

Run the included test script to verify the server works with Claude:

node test-fraud-mcp.js

This script connects to your local MCP server and runs sample fraud analysis scenarios.

Build & Deploy

# Build TypeScript
npm run build

# Start production server
npm start

License

MIT