Cursor Integration

This document describes how to integrate Foggy MCP Service with Cursor IDE.

Configuration Methods

Cursor supports adding MCP servers through settings interface or configuration files.

Method 1: Settings Interface (Recommended)

1. Open Cursor
2. Go to Settings → Features → MCP Servers
3. Click Add Server
4. Fill in configuration:
   • Name: foggy-dataset
   • URL: http://localhost:7108/mcp/analyst/rpc

Method 2: Configuration File

Edit Cursor configuration file:

• macOS: ~/Library/Application Support/Cursor/User/settings.json
• Windows: %APPDATA%\Cursor\User\settings.json
• Linux: ~/.config/Cursor/User/settings.json

Add the following configuration:

{
  "mcp.servers": {
    "foggy-dataset": {
      "url": "http://localhost:7108/mcp/analyst/rpc"
    }
  }
}

Endpoint Selection

Choose endpoints based on use case:

Endpoint	Use Case	Description
/mcp/analyst/rpc	Development & Debug	Structured queries, precise control
/mcp/admin/rpc	Full Access	Includes all tools
/mcp/business/rpc	Simple Queries	Natural language only

Use Cases

Database Query Assistance

During development, have Claude help with:

Query order statistics for the last 7 days from orders table

Data Analysis

Analyze sales data, find top 10 categories by sales amount

Generate Reports

Generate this month's sales comparison chart across stores

Usage in Code

Cursor's AI assistant can call MCP tools during code editing:

// Have Claude query data and generate code
// Input: "Query user table field structure, generate TypeScript interface"

// Claude will call dataset.describe_model_internal to get metadata
// Then generate code like:

interface User {
  id: number;
  username: string;
  email: string;
  createdAt: Date;
}

Configuration with Authentication

If MCP service requires authentication:

{
  "mcp.servers": {
    "foggy-dataset": {
      "url": "http://localhost:7108/mcp/analyst/rpc",
      "headers": {
        "Authorization": "Bearer your-token"
      }
    }
  }
}

Multi-Project Configuration

Configure different MCP services for different projects:

{
  "mcp.servers": {
    "sales-data": {
      "url": "http://sales-server:7108/mcp/analyst/rpc"
    },
    "inventory-data": {
      "url": "http://inventory-server:7108/mcp/analyst/rpc"
    }
  }
}

Verify Configuration

After configuration:

1. Restart Cursor
2. Open AI Chat (Cmd/Ctrl + L)
3. Input: List available semantic layer models
4. Confirm Claude can call dataset.get_metadata tool

Troubleshooting

MCP Tools Not Available

1. Check if MCP service is running:

   curl http://localhost:7108/actuator/health

2. Check Cursor logs:

   • macOS: ~/Library/Logs/Cursor/
   • Windows: %APPDATA%\Cursor\logs\

Configuration Not Taking Effect

1. Ensure JSON format is correct
2. Completely quit and restart Cursor
3. Check URL includes full path (including /rpc)

Connection Timeout

If queries timeout, it may be:

• MCP service responding slowly, check service logs
• Network connection issues
• Query data volume too large, try adding filter conditions

Comparison with Claude Desktop

Feature	Cursor	Claude Desktop
Code Integration	✅ Deep Integration	❌ Standalone App
Context Awareness	✅ Current File/Project	❌ Conversation Only
MCP Support	✅	✅
Configuration Complexity	Low	Low

Next Steps

• Claude Desktop Integration - Use in Claude Desktop
• API Usage - Direct MCP API calls
• Tools List - View all available tools